#Audited: Social Media and Tax Enforcement

Washington and Lee University School of Law Washington and Lee University School of Law

Washington and Lee University School of Law Scholarly Commons Washington and Lee University School of Law Scholarly Commons

Scholarly Articles Faculty Scholarship

2021

#Audited: Social Media and Tax Enforcement #Audited: Social Media and Tax Enforcement

Michelle Lyon Drumbl

Washington and Lee University School of Law

, [email protected]

Follow this and additional works at: https://scholarlycommons.law.wlu.edu/wlufac

Part of the Privacy Law Commons, Science and Technology Law Commons, Taxation-Federal

Commons, and the Tax Law Commons

Recommended Citation Recommended Citation

Michelle Lyon Drumbl, #Audited: Social Media and Tax Enforcement, 99 Or. L. Rev. 301 (2021).

This Article is brought to you for free and open access by the Faculty Scholarship at Washington and Lee University

School of Law Scholarly Commons. It has been accepted for inclusion in Scholarly Articles by an authorized

administrator of Washington and Lee University School of Law Scholarly Commons. For more information, please

contact [email protected].

[301]

Articles

MICHELLE LYON DRUMBL

#Audited: Social Media and Tax

Enforcement

Introduction ...................................................................................... 302

I. Tensions Arising from the Collision of Automation,

Convenience, Privacy, and Expectations .............................. 307

A. Is Our Collective Notion of Privacy Slowly Changing?

Examples Outside the Realm of Tax Administration .... 308

1. Government Agency Use of Social Media Mining

and Big Data ............................................................. 309

2. Private Actor Use of Social Media Mining and

Big Data ................................................................... 318

B. What Is a Taxpayer’s Right to Privacy? ........................ 320

1. IRS Use of Data Analytics—Past and Future .......... 323

2. Disproportionate Enforcement on Lower-Income

Taxpayers, and Implications for Taxpayer Privacy . 326

3. Differing Policy Implications for the IRS

Examination and Collection Functions .................... 328

Michelle Lyon Drumbl, Robert O. Bentley Professor of Law and Tax Clinic Director,

Washington and Lee University School of Law. I wish to extend special thanks to the

participants of the 2019 National Tax Association Annual Conference on Taxation,

including Leandra Lederman, Emily Satterthwaite, Dave Williams, and Ted Afield, and to

Steven Dean, Heather Field, Adam Thimmesch, Miranda Fleisher, Susan Morse, Jordan

Barry, Shu-Yi Oei, Ben Leff, Emily Cauble, Brian Galle, and Darien Shanske for their

comments. Thank you also to Franklin Runge, Mark Drumbl, Jeff Lyon, Leanne Scott, and

Margaret Hu for early stage inspiration and brainstorming. I am incredibly grateful to Seth

Kuntz, Angela Cannon, and Hunter Rush for their stellar research assistance, and to the

Frances Lewis Law Center at the Washington and Lee University School of Law for support

of the project.

302 OREGON LAW REVIEW [Vol. 99, 301

4. Differing Policy Implications for Civil Tax

Enforcement and Criminal Tax Enforcement .......... 332

II. Punishing the Unsophisticated: Pondering Braggadocio,

Whistleblowers, and the Quest to Close the Tax Gap ........... 333

A. The Temptation to Disregard Individual Privacy

Concerns ........................................................................ 334

B. Whistleblowers: Should the IRS Encourage Social

Media Snitching, and Under What Circumstances? ...... 340

C. Proposals for Setting IRS Policies on Social Media

Mining: Balancing Modern Enforcement Techniques

with a Taxpayer Right to Privacy .................................. 343

1. The IRS Should Clarify Its Understanding of the

Taxpayer “Right to Privacy” .................................... 343

2. Increase Transparency of Audit Techniques ............ 344

3. Limit Social Media Investigations to Manual

Searches Rather than Automated, and Define

Limits in the Internal Revenue Manual .................... 344

4. If Automated Social Media Mining Is Used,

Implement Use of Pre-Examination Soft Letters to

Nudge Taxpayers Detected by the Algorithm .......... 347

5. Sharply Define the Social Media Mining Criteria,

Using It to Target Only the Most Egregious

Noncompliance ........................................................ 348

6. Use Social Media Mining Only at Taxpayer’s

Request, as a Method of Dispute Resolution ........... 350

III. Broader Implications for Representing Low-Income

Taxpayers in the #TMI Era ................................................... 351

Conclusion ....................................................................................... 357

NTRODUCTION

ith budget constraints and a mission that has been expanded by

Congress over time, it is not surprising that the Internal Revenue

Service (IRS) is looking for new tools to maximize its enforcement

efficiency. Ever-advancing technology provides new opportunities for

the IRS, and in turn, new privacy concerns for taxpayers.

In December 2018, the IRS made headlines when it issued a request

for information (RFI) seeking social media research tools available in

2021] #Audited: Social Media and Tax Enforcement 303

the marketplace.

In its RFI, the IRS referenced the limits on its own

employees’ abilities to engage in social media research,

and it stated

its hope to engage a vendor-supplied tool that would allow the agency

to access publicly available social media to “expedite IRS case

resolution for existing compliance cases, providing a more efficient

way of identifying resources and assisting with the collection of known

tax deficiencies, leading to increased collection of revenue involving

unfiled tax returns and other tax liabilities.”

The RFI noted that the

agency “respects taxpayer rights” and stated that “such a tool would

not be used to search the internet or social media sites for purposes of

identifying or initiating new tax audits” but rather to “assist with

previously identified tax compliance cases.”

The RFI also mentions a potential benefit to taxpayers, which is that

the IRS believes a social media mining tool could aid in the resolution

of tax-related identity theft.

The RFI does not elaborate further on this

potential use, though it is worth imagining how the IRS might use

social media constructively rather than for enforcement.

The RFI

concludes by noting that the IRS intends to “be mindful that frequently

information posted on social media and the internet may be wrong or

misleading.”

The RFI does not explicitly define the term “social media” or list

examples of the sites the IRS is interested in reviewing. Presumably the

term would encompass websites in which users create profiles and may

interact with one another, such as Facebook, Twitter, Instagram, and

LinkedIn.

Demographic studies reveal that a majority of American

Jared Gilmour, The IRS Wants Help Scouring Social Media for Clues on Tax Cheats,

MIAMI HERALD (Dec. 28, 2018, 4:29 PM), https://www.miamiherald.com/news/nation

-world/national/article223681430.html.

The RFI cites to IRM 11.3.21.8.1(4) as prohibiting IRS employees from logging into

social media sites to carry out compliance-related work and states that “the IRS currently

has no formal tool to access [] public information [used by taxpayers to advertise, promote

and sell products and services], compile social media feeds, or search multiple social media

sites.” DEP’T OF TREASURY, IRS, REQUEST FOR INFORMATION, SOLICITATION NO.

2032H8-RFI-MEDIA,

SOCIAL MEDIA RESEARCH REQUEST (Dec. 18, 2018) [hereinafter

IRS RFI].

Id.

See discussion infra Section II.C.6.

IRS RFI, supra note 2; see also infra Section II.A (discussing the seeming contradiction

between the IRS’s goals of efficiency and the extra cost and work that would be required

for a human being to sort through misleading claims).

See Aaron Smith & Monica Anderson, Social Media Use in 2018, PEW RSCH.

CTR. (Mar. 1, 2018), https://www.pewresearch.org/internet/2018/03/01/social-media-use

304 OREGON LAW REVIEW [Vol. 99, 301

adults use Facebook and YouTube on their computer or cellphone.

Younger adults (age 18–24) use a wider variety of sites and do so more

frequently.

A 2012 study showed that younger adults are more likely

to use social media than older adults, and urban dwellers are

significantly more likely to use social media than those who live in

rural areas; however, the study did not reveal a statistically significant

disparity in overall social media site use across different household

income levels.

Interestingly, even those who abstain from social

media may not be as hidden from sight as they think. Researchers report

that individuals with no social media accounts themselves are likely

drawn into this public sphere, with algorithms able to draw predictive

and reasonably accurate

findings about an individual based upon

postings by those of the individual’s friends, family, and acquaintances

who do have a social media presence.

The thought of IRS employees searching through individual

taxpayers’ social media sites presents concerns on multiple fronts:

What policies or limits might the agency adopt? Might the IRS later

expand the program beyond the stated intention to use such practices

only in existing compliance cases? This Article begins by considering

the potential implications to taxpayer privacy rights of such an

initiative, and questions whether it may conflict with the statutory

notion of a taxpayer’s right to privacy.

Speaking for myself, I bristle

-in-2018/ [https://perma.cc/L432-JGTW]. Smith and Anderson also define social media as

including YouTube, Snapchat, Pinterest, and WhatsApp. Id.

Id. (noting that 68% of American adults are Facebook users, with 74% of Facebook

users accessing it on a daily basis, and that a majority of American adults in all demographic

groups, other than individuals 65 and older, currently use Facebook).

Id.

Maeve Duggan & Joanna Brenner, The Demographics of Social Media Users — 2012,

PEW RSCH. CTR. 3 (Feb. 14, 2013), https://www.pewresearch.org/internet/2013/02/14/the

-demographics-of-social-media-users-2012/ [https://perma.cc/U8AX-8NXN].

Jessica Baron, Think Your Data Is Private Because You’re Not on Social Media?

Think Again, FORBES (Jan. 23, 2019, 8:46 AM), https://www.forbes.com/sites/jessicabaron

/2019/01/23/think-your-data-is-private-because-youre-not-on-social-media-think-again

/#7ab38b814a16 [https://perma.cc/HL8C-2BQW] (“The original user’s Tweets allowe

them to predict future tweets with an accuracy rate of roughly 64% and the user’s contacts

gave them enough data to predict behavior with an accuracy rate of 61%.”).

Id. Baron’s article describes the work of James P. Bagrow et al., Information Flow

Reveals Prediction Limits in Online Social Activity, 3 NAT. HUM. BEHAV. 122, 122 (2019)

(“As few as 8–9 of an individual’s contacts are sufficient to obtain predictability comparable

to that of the individual alone.”).

Concerns about taxpayer privacy are certainly not new. Tax scholars such as Michael

Hatfield have written multiple thoughtful pieces addressing privacy in taxation, including

the scope of the IRS’s power to collect personal information. See Michael Hatfield, Privacy

in Taxation, 44 F

LA. ST. U. L. REV. 579, 581 (2017) (“[A]mong government agencies, the

2021] #Audited: Social Media and Tax Enforcement 305

at the idea of the IRS staking out an ever-more invasive footprint into

the private life of any individual taxpayer. Regardless of one’s views

on taxpayer privacy, this Article makes a deeper inquiry: it asks not

whether the IRS has the right to access and use this information for

examination and collections purposes but whether, as a matter of social

policy, it is appropriate and desirable for the IRS to trawl for such

information, and to do so in a civil context as opposed to in a criminal

case.

Beyond questions of privacy, a broader concern is the potential for

social media mining to create disproportionate harms for low-income

taxpayers. As this Article will describe, low-income taxpayers are

already subject to disproportionate rates of tax enforcement relative to

most other income bands. Moreover, this economically vulnerable

population is also subject to intrusive and judgmental monitoring in

other contexts.

On balance, it strikes me that to use social media

mining as a tax enforcement tool is to simply add a layer of further

indignity onto a population that is already subject to increased digital

surveillance by virtue of lacking income or wealth.

Thus, the primary focus of this Article is to question whether, in light

of the IRS’s need to make the most of diminished resources,

it is

equitable to further automate its examination and collections tactics in

a way that punishes unsophisticated behavior. In particular, this Article

articulates a concern that the use of social media mining may pose a

greater harm to low-income taxpayers relative to other types of

taxpayers, in part because it is easier for the IRS to direct automated

resources at the types of issues involved in examining those returns.

Many low-income families rely on the tax filing system to claim critical

social welfare benefits such as the earned income tax credit and child

IRS has the broadest legal authority to collect information that minimizes privacy . . . .”).

Likewise, Kimberly Houser and Debra Sanders have raised concerns about how the IRS is

already using social media to collect taxpayer information, asserting that the IRS is engaging

in social media data mining in ways that breach taxpayer privacy. Kimberly A. Houser &

Debra Sanders, The Use of Big Data Analytics by the IRS: Efficient Solutions or the End of

Privacy as We Know It?, 19 V

AND. J. ENT. & TECH. L. 817, 817 (2017).

See infra notes 30–33 and accompanying text.

See, e.g., IRS Oversight: Treasury Inspector General for Tax Administration: Hearing

Before the Subcomm. on Fin. Servs. & Gen. Gov’t of the H. Comm. on Appropriations, 116th

Cong. 86 (2019) (testimony of J. Russell George, Treasury Inspector General for Tax

Administration) [hereinafter George testimony]; Robert A. Weinberger, Budget Blues for

Tax Administration, T

AX POL’Y CTR. (Jan. 13, 2020) (showing in a chart how IRS

appropriations fell by 21% from 2010 to 2020), https://www.taxpolicycenter.org/taxvox

/budget-blues-tax-administration [https://perma.cc/G6ML-U2UY].

See infra Section I.B.2.

306 OREGON LAW REVIEW [Vol. 99, 301

tax credit. Congress has chosen this system to reward and incentivize

work and to help lift working families from poverty.

In many cases,

the working families eligible for these benefits have a zero-tax liability

because they earn less than the standard deduction. While taxpayers

with annual income below the standard deduction are generally not

required to file an income tax return, they must file a return to claim

these valuable refundable credits. Concerns about how enforcement

techniques involving social media mining may disproportionately

impact low-income taxpayers are especially relevant in light of data

from recent years showing that the IRS audits the poor at a similar rate

to the highest-earning taxpayers

and in light of the IRS defending that

practice as justified given its available resources.

To clarify, this Article does not condone tax noncompliance, and

as a tax attorney, I would never advise my clients that “a little cheating

is OK.” While taxpayer privacy and the possibility of disparate

treatment of taxpayers are the main concerns of this Article, my

thoughts are tempered with realism and cost-benefit considerations.

Just as the police tolerate a little bit of speeding (e.g., ignoring those

who drive seventy-four miles per hour in a seventy mile per hour zone,

but pursuing those who drive eighty-five miles per hour), it does not

seem practical or cost efficient to pursue every perceived case of tax

See, e.g., MICHELLE LYON DRUMBL, TAX CREDITS FOR THE WORKING POOR: A

CALL FOR REFORM (2019) (detailing the history of the earned income tax credit and child

tax credit and exploring the reasons why Congress has chosen the IRS to administer these

social benefits).

Paul Kiel, It’s Getting Worse: The IRS Now Audits Poor Americans at About the Same

Rate as the Top 1%, PROPUBLICA (May 30, 2019, 10:16 AM), https://www.propublica.org

/article/irs-now-audits-poor-americans-at-about-the-same-rate-as-the-top-1-percent [https:

//perma.cc/CAK2-SE25].

William Hoffman, IRS Exams Focus on EITC Claims, Not Poor, Inspector General

Says, TAX NOTES TODAY (Sept. 27, 2019), https://www.taxnotes.com/tax-notes-today

-federal/tax-system-administration/irs-exams-focus-eitc-claims-not-poor-inspector-general

-says/2019/09/27/29zm4 [https://perma.cc/G4RK-P9WE]. Subcommittee Chair Mike

Quigley asked, “So if you have fewer resources, it makes sense to audit poor people more?”

In response, Treasury Inspector General J. Russell George replied, “It’s somewhat more

complicated than that” and explained that audit work on EITC claimants is easier than on

high-income taxpayers, “especially with the work of junior IRS employees. . . . The more

sophisticated the income tax, the more involved it is, the longer it takes. It really boils down

to how [the IRS] allocates their resources.” Id. (alteration in original). Hoffman reports that

when asked about possible racism or bias in IRS examination policies, George replied,

“None of our work has shown any evidence that bias is occurring in terms of those with

money versus those with less money . . . [b]ut there is no question that more low-income

people are being examined than upper-income people.” Id.; see also George testimony,

supra note 16, at 86 (“As a percentage of overall enforcement revenue collected, the amoun

attributable to automated [functions are] increasing . . . [while o]ther types of enforcement

actions . . . are decreasing.”).

2021] #Audited: Social Media and Tax Enforcement 307

noncompliance. For example, is it worth it to devote resources to

having the IRS manually investigate every social media post that

suggests any hint of unreported income, no matter how small in dollar

amount?

This Article proceeds in three parts. Part I identifies the growing

trend for public agencies and private third parties to use social media,

automation, and other tools of technology to monitor or collect

information on individuals. It identifies concerns and tensions that arise

from this trend. It then considers how a taxpayer’s statutory “right to

privacy” applies in the context of information made available online by

the taxpayer. While the taxpayer right to privacy is not clearly defined,

it appears to refer to more than just the right of confidentiality of

taxpayer information. Part II frames the issue of social media mining

within broader reactions to punishing (or choosing not to punish)

unsophisticated behavior and raises thoughts about how social media

mining is different than investigating tips from whistleblowers. Part II

then makes several proposals for the IRS to consider, starting with a

call for the agency to provide the public with a clearer explanation of

its understanding of the taxpayer “right to privacy.” Part III, proceeding

under the assumption that the IRS can and will access social media for

enforcement purposes,

provides takeaways and thoughts for those

attorneys who advocate on behalf of low-income taxpayers and

represent them in their controversies with the IRS.

ENSIONS ARISING FROM THE COLLISION OF AUTOMATION,

CONVENIENCE, PRIVACY, AND EXPECTATIONS

Examples abound of federal and state agencies, as well as private

third parties, using big data algorithms, automated systems, and other

tools of technology (including, but not limited to, social media mining)

to monitor or collect information on individuals. In some cases,

This fits within broader questions of the appropriateness of leeway in law

enforcement. See Shu-Yi Oei & Diane Ring, Falling Short in the Data Age (working paper)

(on file with author) (“[I]ncreasing access to data and information will change the

availability and shape of informal leeway in the law.”).

While the primary focus of this Article is the possibility that the IRS may use social

media at the examination and collections level, a recent Tax Court decision reveals the first

known example of IRS Counsel using a social media post as evidence contradicting the

petitioner’s filing position. James Creech, Oversharing on Social Media Reaches the Tax

Court, P

ROCEDURALLY TAXING (Sept. 3, 2020) (citing Brzyski v. Comm’r, T.C. Summ.

Op. 2020-25 (Aug. 27, 2020)), https://procedurallytaxing.com/oversharing-on-social-media

-reaches-the-tax-court/ [https://perma.cc/S99C-3B48].

308 OREGON LAW REVIEW [Vol. 99, 301

individuals consent knowingly as a matter of convenience. In other

instances, individuals are unaware they are being surveilled. The

motivations for tracking individuals and collecting such data are varied,

with rationales ranging from national security to marketing of

consumer products and many things in between.

Section I.A first discusses some of the ways in which government

agencies other than the IRS use technology to categorize or identify

behaviors, and it then provides examples of how private third parties

do so, including as a service to other individuals and, in some cases,

the person being surveilled. Section I.A of the Article raises questions

more than it provides answers: Where are we as a society with respect

to privacy norms, and what is our comfort level with technological

tracking? Under what circumstances do Americans seem to find it

acceptable to sacrifice privacy (whether their own or that of others) as

a means to achieve law enforcement, safety, or convenience? The

examples used herein are intended to set a stage for how the public at

large might feel about the IRS engaging in data mining for tax

enforcement purposes.

Section I.B addresses notions of tax and privacy, reflecting on how

the IRS has historically used publicly available information and

imagining how the use of social media mining may change or build

upon that use. The RFI reveals that the IRS is currently researching

third-party options for machine-driven data mining on publicly

available data.

When asked whether the government was interested

in leveraging artificial intelligence-based technology to interpret text

and image-based information on social media, the IRS responded that

it “is open to any viable solution that fits our needs.”

Thus, this

Article imagines a variety of possibilities that might flow once the IRS

opens the Pandora’s box that is social media.

A. Is Our Collective Notion of Privacy Slowly Changing? Examples

Outside the Realm of Tax Administration

As technology evolves, so too do norms and expectations. Public

agencies and private companies alike use artificial intelligence, or AI,

for a variety of predictive purposes. In some instances, predictive

screening techniques may be employed by both government agencies

and private industry in pursuit of the same goals, either in partnership

IRS RFI, supra note 2.

IRS, REQUEST FOR INFORMATION, NO. 2032H8-RFI-MEDIA, ANSWER 31, IRS

SUPPLIED RESPONSES TO VENDOR QUESTIONS (on file with author).

2021] #Audited: Social Media and Tax Enforcement 309

or in parallel. For example, both the U.S. Department of Veterans

Affairs and Facebook have created predictive tools for suicide

screening and prevention.

The use of artificial intelligence by

government agencies may have different policy or privacy implications

than the use by private companies, but there also may be overlap. With

its RFI, the IRS signals an intention to partner with private industry.

Thus, in thinking about how the IRS may use social media to mine for

data, it is helpful to consider analogies drawn both from other

government agencies and from private companies.

1. Government Agency Use of Social Media Mining and Big Data

Federal and state government agencies engage in data mining in

variety of ways and for a whole host of purposes.

In their article

addressing the IRS big data analytics program, Kimberly Houser and

Debra Sanders reference a Senate report, noting that as of 2007 there

were 52 different federal agencies using at least 199 different operating

or planned data mining programs, including the IRS.

Many of these

programs involved counterterrorism and national security but also

included efforts to improve agency service or performance, analyze

scientific and research information, manage human resources, and

detect fraud, waste, and abuse.

Government data mining fits within, or perhaps couples with, a

larger trend of using automation to facilitate the administration of

Mason Marks, Artificial Intelligence-Based Suicide Prediction, 21 YALE J.L. & TECH.

SPECIAL ISSUE 98, 102 (2019). Marks divides the types of AI tools used for suicide

screening into two categories: “medical suicide prediction,” which is undertaken within the

healthcare system and involves the use of patient medical records, and “social suicide

prediction,” which refers to tech companies scrutinizing data, including social media data.

Id. at 104–05. Medical suicide prevention is subject to various regulatory regimes that

govern the healthcare industry, including but not limited to the Health Insurance Portability

and Accountability Act. Id. at 105. Marks explains that the social suicide prevention is more

controversial in that it lacks accountability and transparency. Id. at 109. As a government

agency, Veterans Affairs is able to engage in both types of categories of suicide prevention

screenings, medical and social. Id. at 105. Tech companies, on the other hand, do not have

access to medical records due to privacy laws, and engage only in social suicide prevention

screenings. Id. at 107.

As Houser and Sanders note, data mining can be predictive (meaning it analyzes and

extrapolates data to make predictions about unavailable data) or descriptive (meaning it

“summarizes properties of the data set”). Houser & Sanders, supra note 14, at 824.

Id. at 825; Balancing Privacy and Security: The Privacy Implications of Government

Data Mining Programs: Hearing Before the S. Comm. on the Judiciary, 110th Cong. 1

(2007).

U.S. GEN. ACCT. OFF., GAO 04-548, DATA MINING: FEDERAL EFFORTS COVER A

WIDE RANGE OF USES 8 (2004).

310 OREGON LAW REVIEW [Vol. 99, 301

social services to the public.

While efficient, automation also has

identifiable downsides for individuals. In her book Automating

Inequality, Virginia Eubanks provides alarming examples of how

automated systems, ranking algorithms, and predictive risk models

affect the lives of the poor and working-class communities, including

individuals with respect to their use of various public services.

Eubanks provides three in-depth case studies in her book: automation

of state welfare eligibility, an electronic registry of unhoused

individuals, and a risk model created to predict which children might

become victims of child abuse or neglect.

Eubanks notes that

technologies are being integrated into social services “at a breathtaking

pace, with little or no political discussion about their impacts,” and she

observes ways in which they “intensif[y] discrimination” and

disproportionately affect low-income communities.

Eubanks notes

that this digital trend is a modern continuation of the centuries-old ways

in which the poor and working class are surveilled and stigmatized for

being poor; she refers to this as the “digital poorhouse.”

Other scholars have identified similar concerns about the link

between poverty and privacy. Using empirical data from a survey of

low-income individuals, Mary Madden, Michele Gilman, Karen Levy,

and Alice Marwick developed case studies of how big data tracking

(including social media tracking) poses risks to the poor in the context

of employment, access to higher education, and policing.

These observations by Eubanks, as well as Madden and her

coauthors, reinforce my concern that low-income taxpayers may find

themselves under disproportionate scrutiny relative to other taxpayers.

As I describe in Section I.B, the IRS’s use of automated examination

For a discussion envisioning how regulators might use big data to personalize or

individualize outcomes, see Jordan M. Barry, John William Hatfield & Scott Duke

Kominers, To Thine Own Self Be True? Incentive Problems in Personalized Law, 62 W

& MARY L. REV. (forthcoming 2021).

VIRGINIA EUBANKS, AUTOMATING INEQUALITY: HOW HIGH-TECH TOOLS PROFILE,

POLICE, AND PUNISH THE POOR 9 (2018) (“Digital tracking and decision-making systems

have become routine in policing, political forecasting, marketing, credit reporting, criminal

sentencing, business management, finance, and the administration of public programs.”).

Id. at 10.

Id. at 11–12.

Id. at 12.

Mary Madden, Michele Gilman, Karen Levy & Alice Marwick, Privacy, Poverty, and

Big Data: A Matrix of Vulnerability for Poor Americans, 95 WASH. U. L. REV. 53, 54

(2017). “In each setting, low-income Americans face not only adverse inferences drawn

based on their personally identifiable information (which often is erroneous), but also those

drawn from their social media and demographic networks.” Id. at 124.

2021] #Audited: Social Media and Tax Enforcement 311

techniques already disproportionately impacts the low-income

taxpayer population.

Further exacerbating my concern, social science

research has found that lower-income individuals are “comparatively

unlikely to restrict their social media activity’s visibility via privacy

settings and are less hesitant to share sensitive information online.”

Thus the very nature of their social media use makes them vulnerable

to disproportionate scrutiny.

Inherent in concerns about the low-income populations is a concern

about how these enforcement practices may impact taxpayers of

different races. Scholars such as Dorothy Brown, Palma Joy Strand,

Nicholas Mirkay, and Francine Lipman have identified structural ways

in which income inequality and wealth inequality are racialized and

how tax structures perpetuate or exacerbate these inequalities.

Ideally, the use of artificial intelligence and technology would be used

to reduce inequality.

However, much like humans, algorithms are

susceptible to bias.

Data privacy scholar Margaret Hu has argued that

“algorithmically anchored” screening protocols are reminiscent of Jim

Crow laws: the protocols may present on the front end as facially

neutral while the results “may in fact have a disparate impact on

traditionally protected classes.”

Information studies scholar Safiya

I have also addressed this phenomenon in my previous scholarship. See Michelle Lyon

Drumbl, Those Who Know, Those Who Don’t, and Those Who Know Better: Balancing

Complexity, Sophistication, and Accuracy on Tax Returns, 11 PITT. TAX REV. 113, 135

(2013) (examining and rethinking the application of the accuracy-related penalty to

unsophisticated and low-income taxpayers, particularly in the context of complex provisions

granting social benefits, such as the EITC).

Spencer Headworth, Getting to Know You: Welfare Fraud Investigation and the

Appropriation of Social Ties, 84 A

M. SOC. REV. 171, 189 (2019) (citing Madden et al., supra

note 34).

DOROTHY BROWN, THE WHITENESS OF WEALTH 19–21 (2021) (describing how the

persistent and widening black-white wealth gap is related to tax policy); see also Palma Joy

Strand & Nicholas A. Mirkay, Racialized Tax Inequity: Wealth, Racism, and the U.S. System

of Taxation, 15 N

W. J.L. & SOC. POL’Y 265, 265 (2020); Francine J. Lipman, Nicholas A.

Mirkay & Palma Joy Strand, U.S. Tax Systems Need Anti-Racist Restructuring, 168 T

NOTES FED., Aug. 3, 2020, at 855, 856.

See, e.g., I. Bennett Capers, Race, Policing, and Technology, 95 N.C. L. REV. 1241,

1291 (2017) (positing that technology can be harnessed in ways that would deracialize

policing, or at least make a “significant step in the right direction.”).

Rumman Chowdhury & Narenda Mulani, Auditing Algorithms for Bias, HARV. BUS.

REV., Oct. 24, 2018; see also Kristian Lum, Limitations of Mitigating Judicial Bias with

Machine Learning, 1 N

AT. HUM. BEHAV., No. 0141, June 26, 2017 (noting that judicial use

of risk assessment tools in criminal cases are only as objective as how they are trained

because machines will absorb the underlying human biases inherent in the data inputs),

https://www.nature.com/articles/s41562-017-0141.pdf [https://perma.cc/5C2H-DHAZ].

Margaret Hu, Algorithmic Jim Crow, 86 FORDHAM L. REV. 633, 645 (2017). Hu

writes,

312 OREGON LAW REVIEW [Vol. 99, 301

Umoja Noble’s recent book, Algorithms of Oppression, provides

compelling examples of how algorithmic search engines discriminate

against people of color, particularly women, and reinforce negative

race and gender stereotypes.

Privacy experts express concern over many different ways in which

government agencies use cutting-edge technology and artificial

intelligence—not just social media—as surveillance tools,

and the

potential negative consequences to the public.

As one example,

without knowledge or consent, the Federal Bureau of Investigation

and the Immigration and Customs Enforcement agencies are using

facial recognition software to scan state driver’s license databases as a

“routine investigative tool.”

While the use of this has grown over

time, a 2019 report revealed that facial recognition software

Artificial intelligence and algorithms are not usually perceived as resulting in

discrimination. In fact, they may appear to be equality-compliant or even equality-

enhancing in that algorithmic screening and vetting can be applied equally across

entire populations and subpopulations. Screening and classification systems,

however, even when facially neutral and algorithmically based, can lead to

profound constitutional challenges.

Id. at 650.

SAFIYA UMOJA NOBLE, ALGORITHMS OF OPPRESSION: HOW SEARCH ENGINES

REINFORCE RACISM 17–18 (2018). As one of many examples, Noble describes her surprise

when a Google search on her own computer using the term “black girls” retrieved results

filled with pornography, despite the fact that her individual search history included

engagement with Black feminist texts, videos, and books. Id.

Margaret Hu, Big Data Blacklisting, 67 FLA. L. REV. 1736, 1744 (2015) (explaining

how certain government programs screen the public through big data protocols to create a

class of big data blacklisted individuals) (“Specifically, [Hu’s article] explains how, for

example, matches and mismatches in big data systems can lead to inferential guilt that can

directly or indirectly categorize individuals as administratively ‘guilty until proven

innocent’ by virtue of digitally generated suspicion.”).

Of airport facial scanning for security screening, technology columnist Geoffrey

A. Fowler writes, “This has all the makings of a convenience trap. That’s how privacy-

invading technology—the stuff of China’s police state—creeps into American life.

”

Geoffrey A. Fowler, Don’t Smile for Surveillance: Why Airport Face Scans Are a Privacy

Trap, W

ASH. POST (June 10, 2019, 1:51 PM), https://www.washingtonpost.com/technology

/2019/06/10/your-face-is-now-your-boarding-pass-thats-problem/ [https://perma.cc/B63F

-MS44]; see also Shea Swauger, Software That Monitors Students During Tests Perpetuates

Inequality and Violates Their Privacy, MIT

TECH. REV. (Aug. 7, 2020), https://www

.technologyreview.com/2020/08/07/1006132/software-algorithms-proctoring-online-tests

-ai-ethics/ [https://perma.cc/R8BB-QWZC] (describing how algorithmic proctoring of

online exams, which includes a facial recognition component, can result in racial and gender

biases).

Drew Harwell, FBI, ICE Find State Driver’s License Photos Are a Gold Mine

for Facial-Recognition Searches, WASH. POST (July 7, 2019, 12:54 PM), https://www

.washingtonpost.com/technology/2019/07/07/fbi-ice-find-state-drivers-license-photos-are

-gold-mine-facial-recognition-searches/ [https://perma.cc/Z3KL-4R2V].

2021] #Audited: Social Media and Tax Enforcement 313

misidentified people of color more frequently than white people and

misidentified women more frequently than men.

In what has been

described as a first known case (and a terrifyingly dystopian one at

that), a Black man in Detroit was wrongfully arrested and detained for

thirty hours based on an erroneous match from a facial recognition

algorithm.

Ironically, social media could have helped in this case: the

man later realized he could have used his Instagram account to provide

an alibi by showing he was elsewhere at the time the crime was

committed.

As a result of concerns about discrimination and the

potential for false positives, a handful of localities have banned the use

of such technology.

In the wake of nationwide protests of police

brutality following the killing of George Floyd, Amazon announced a

one-year moratorium on the police use of Rekognition, the company’s

facial recognition technology,

and IBM announced that it “firmly

opposes and will not condone” the use of such technology “for mass

Drew Harwell, Federal Study Confirms Racial Bias of Many Facial-Recognition

Systems, Casts Doubt on Their Expanding Use, WASH. POST (Dec. 19, 2019, 3:43 PM),

https://www.washingtonpost.com/technology/2019/12/19/federal-study-confirms-racial

-bias-many-facial-recognition-systems-casts-doubt-their-expanding-use/ [https://perma.cc

/B3CR-DM5X] (“The National Institute of Standards and Technology, the federa

laboratory known as NIST that develops standards for new technology, found ‘empirical

evidence’ that most of the facial-recognition algorithms exhibit ‘demographic differentials’

that can worsen their accuracy based on a person’s age, gender or race.”); see also Natasha

Singer & Cade Metz, Many Facial-Recognition Systems Are Biased, Says U.S. Study, N.Y.

IMES (Dec. 19, 2019), https://www.nytimes.com/2019/12/19/technology/facial-recognition

-bias.html [https://perma.cc/4QB4-69FL] (“The systems falsely identified African-

American and Asian faces 10 times to 100 times more than Caucasian faces . . . [a]nd falsely

identified older adults up to 10 times more than middle-aged adults.”).

Kashmir Hill, Wrongfully Accused by an Algorithm, N.Y. TIMES (Aug. 3, 2020), https:

//www.nytimes.com/2020/06/24/technology/facial-recognition-arrest.html [https://perma

.cc/L9D3-TDZZ]. While referring to this as the first known case of its kind, the article quotes

Clare Garvie, a lawyer from the Georgetown University Center on Privacy and Technology

who has written about concerns with facial recognition technology: “I strongly suspect this

is not the first case to misidentify someone to arrest them for a crime they didn’t commit.

This is just the first time we know about it.” Id.

Id. In the proposals I set forth in Section II.C, I suggest the IRS might restrict its use

of social media mining to constructive purposes, such as when the taxpayer requests it to

resolve a dispute.

Harwell, supra note 45 (stating that San Francisco and Oakland in California, and

Somerville and Brookline in Massachusetts, passed bans in 2019 on facial recognition use

by public officials, and the State of California banned the software’s use in police body

cameras).

Jay Greene, Amazon Bans Police Use of Its Facial-Recognition Technology for a

Year, W

ASH. POST (June 10, 2020, 5:31 PM), https://www.washingtonpost.com/technology

/2020/06/10/amazon-rekognition-police/ [https://perma.cc/S38P-A2K8].

314 OREGON LAW REVIEW [Vol. 99, 301

surveillance, racial profiling, [or] violations of basic human rights and

freedoms.”

As to the appropriateness of government agencies using big data

mining techniques, one might also draw a logistical distinction between

civil and criminal investigations. One might more readily support the

use of artificial intelligence, social media mining, and digital

surveillance to investigate a crime, and I address this distinction in

Section I.B. In some cases, a warrant may be required as a procedural

safeguard prior to a search, but Fourth Amendment protections

generally do not extend to government agents gathering information on

social media.

In this regard, social media use is similar to the

countless ways in which we are all being monitored daily without a

warrant: by security cameras in stores, by red-light cameras,

and by

email providers and internet servers.

In many respects, we consensually (if unthinkingly) sacrifice our

own privacy any time we leave the house.

How, then, can anyone

reasonably expect to have privacy rights in any of the information they

post on Twitter, Instagram, or Facebook? It is true that those platforms

offer a sliding continuum of privacy options; for example, one can

choose to share Facebook posts only with those they have “friended”

on Facebook. That said, we obviously lose control of information the

instant we publish it even to a limited audience; for example, most

people are cognizant that a private email or text sent to a friend, or a

screenshot thereof, can be easily forwarded to an unlimited number of

BBC News, IBM Abandons ‘Biased’ Facial Recognition Tech, BBC (June 9, 2020),

https://www.bbc.com/news/technology-52978191 [https://perma.cc/6QZ6-K75R].

See, e.g., United States v. Gatson, No. 13-705, 2014 WL 7182275, at *22 (D.N.J. Dec.

16, 2014) (holding that an undercover operation in which law enforcement created an

undercover Instagram account and “friended” the suspect in order to view photos and other

information that he posted to his account did not require a warrant). For a comprehensive

discussion of privacy rights and social media posts, including a summary of Gatson and

other relevant caselaw, see Brian Mund, Social Media Searches and the Reasonable

Expectation of Privacy, 19 Y

ALE J.L. & TECH. 238, 240 (2018) (“An exploration of the

extant case law shows that social media users have no reasonable expectation of privacy in

their social media postings—even if users communicate their information behind password-

protected pages.”).

I. Bennett Capers cites red-light cameras and other technological innovations, coupled

with access to big data, as other possible ways to make “policing more transparent,

accountable, and egalitarian.” I. Bennet Capers, Techno-Policing, 15 O

HIO ST. J. CRIM. L.

495, 499 (2018).

See Richard A. Posner, Privacy, Surveillance, and Law, 75 U. CHI. L. REV. 245, 248

(2008) (“[For example,] a person would have to be a hermit to be able to function in our

society without voluntarily disclosing a vast amount of personal information to a vast array

of public and private demanders.”).

2021] #Audited: Social Media and Tax Enforcement 315

other people. And people, even those not on social media themselves,

have no control at all over what others may post about them.

In comparing the IRS to other government agencies, perhaps the

most useful analogy would be to those agencies that administer social

welfare benefits, such as the Social Security Administration and state

departments of social services. I draw upon these examples in particular

because Congress has chosen to use the IRS, and the tax filing process,

to administer and deliver a number of social benefits to the public.

Most notable of these benefits is the earned income tax credit (EITC),

a refundable tax credit paid to approximately 25 million low-income

families each year as part of their income tax refund.

Lawrence

Zelenak has aptly described the EITC as “a welfare program that

happens to be administered through the tax system”

while also

suggesting that because it is housed in the Internal Revenue Code, there

is a higher tolerance for EITC overpayments than for general welfare

overpayments.

Unlike the EITC, for which claimants self-certify, more traditional

welfare benefits such as the Supplemental Nutrition Assistance

Program (SNAP),

Temporary Assistance for Needy Families

(TANF), and Social Security Disability Insurance (SSDI) are subject to

verification procedures before the benefits are awarded. In addition,

recipients of these benefits are also subject to investigation if fraud is

suspected during or after receipt of the benefit.

This use of social media mining for investigating welfare fraud

appears to be widespread among state and local agencies, in particular

Over time, these benefits have included income-based refundable credits to working

families, as well as expenditure-based refundable credits such as the First-Time Homebuyer

Credit, the Adoption Tax Credit, the Premium Tax Credit, and a variety of education credits.

Drumbl, Those Who Know, supra note 35, at 119–39.

EITC Fast Facts, IRS, https://www.eitc.irs.gov/partner-toolkit/basic-marketing

-communication-materials/eitc-fast-facts/eitc-fast-facts [https://perma.cc/MC72-CADU].

Lawrence Zelenak, Tax or Welfare? The Administration of the Earned Income Credit,

52 UCLA L. REV. 1867, 1869 (2005). Zelenak distinguishes the EITC from other welfare

programs, such as SNAP and TANF, because the EITC is predicated on earned income.

Id. at 1874.

SNAP is the program formerly known as Food Stamps. U.S. GOV’T ACCT. OFF., GAO

14-641, SUPPLEMENTAL NUTRITION ASSISTANCE PROGRAM: ENHANCED DETECTION

TOOLS AND REPORTING COULD IMPROVE EFFORTS TO COMBAT RECIPIENT FRAUD, at 3

(2014) [hereinafter SUPPLEMENTAL NUTRITION ASSISTANCE PROGRAM].

See Headworth, supra note 36, at 171 (“[F]ederal law requires state governments to

maintain dedicated fraud control units . . . . These units are busy [and] primarily focus [on

SNAP.]”).

316 OREGON LAW REVIEW [Vol. 99, 301

for enforcement of SNAP eligibility.

The Social Security

Administration also has publicly acknowledged that its adjudicators

use social media in evaluating cases and has stated it will consider

expanding such use.

Sociologist Spencer Headworth describes how SNAP investigators

manually search social media by viewing publicly available posts to

gather information about household circumstances, such as who lives

in the household, relationship status, and vacation photos.

Some

fraud investigations arise from social media posts that are less subtle:

Washington State Department of Social and Health Services fraud

investigators came across a social media post offering to trade “great

organic marijuana” for an EBT card, leading to an arrest of an

individual on both food benefit trafficking and drug charges.

In other

cases, social media leads originate from outside the agency: in

Pennsylvania, a woman who offered on Facebook to trade her EBT

card for cash was arrested after a member of the public tipped off the

state’s fraud unit.

See, e.g., SUPPLEMENTAL NUTRITION ASSISTANCE PROGRAM, supra note 58

(reviewing fraud detection tools used in eleven states, including both automated and manual

monitoring of social media).

SOC. SEC. ADMIN., BUDGET ESTIMATES AND RELATED INFORMATION: FY 2020,

https://www.ssa.gov/budget/FY20Files/2020BO_1.pdf [https://perma.cc/UH24-TCDY].

The Budget Overview states,

Currently, agency adjudicators use social media information to evaluate a

beneficiary’s symptoms when there is a [Cooperative Disability Investigation]

unit’s Report of Investigation that contains social media data corroborating the

investigative findings. In FY 2019, we are evaluating how social media could be

used by disability adjudicators in assessing the consistency and supportability of

evidence in a claimant’s case file.

Id. at 26; see also Sarah Min, Social Security May Use Your Facebook and Instagram Photos

to Nix Disability Claims, CBS NEWS (Mar. 21, 2019, 8:46 AM), https://www.cbsnews.com

/news/social-security-disability-benefits-your-facebook-instagram-posts-could-affect-your

-social-security-disability-claim/ [https://perma.cc/R34V-6VFJ].

Headworth, supra note 36, at 187–88. Of his interviews with fraud enforcement

workers, “[s]ome investigators described Facebook as their most valuable tool, and some

saw it as much less useful.” Id. at 188.

Press Release, Want to Trade EBT Card for Marijuana?, WASH. ST. DEP’T OF SOC.

& HEALTH SERVS. (Feb. 6, 2017), https://www.dshs.wa.gov/sesa/office-communications

/media-release/want-trade-ebt-card-marijuana [https://perma.cc/55U5-U8N4]. Journalis

Kalena Thomhave took note of the way in which this specific press release “celebrated the

arrest” and how the press release “began ominously, ‘Note to would-be food benefits

traffickers. You’re being watched.’” Kalena Thomhave, Another Way to Police the Poor

AM. PROSPECT (Mar. 11, 2019), https://prospect.org/economy/another-way-police-poor/

[https://perma.cc/RZV9-RCFV].

Press Release, OIG Charges Williamsport Woman with SNAP Trafficking for Selling

Her Food Stamp Benefits on Facebook. She Exchanged Benefits for Heroin, PA. OFF. OF

2021] #Audited: Social Media and Tax Enforcement 317

Some state agencies have tried using software programs as an

automated tool for social media mining, with mixed results. The

Department of Agriculture Food and Nutrition Service (FNS) provided

guidance to state agencies that wanted to set up such software to detect

benefit trafficking on social media sites.

A U.S. Government

Accountability Office review of eleven states that experimented with

that approach reported that only one state found the automated

approach effective; the other states found the automated tools

impractical (either because of technical difficulties or because of the

high false positive rate) and preferred to devote their resources to

manual searches.

As public policy student and freelance writer Kalena Thomhave

explains, social media investigations are a modern-day continuation of

the type of invasion of privacy endemic in welfare fraud

investigations.

Thomhave attributes this to the “dichotomy between

those who are deserving and undeserving of public benefits, with the

government as arbiter, [which] lies at the foundation of the American

social safety net.”

Though Thomhave’s remarks are directed at other

agencies that administer social benefits and do not mention the IRS,

one cannot ignore the fact that the EITC is part of the social safety net

for working families and the fact that the IRS has identified the EITC

as a targeted area of enforcement.

This underscores my concern that

the IRS may set its social media sights on EITC recipients, a concern I

address in more detail in Section I.B.

INSPECTOR GEN. (Mar. 24, 2017), https://www.media.pa.gov/Pages/inspector-general

-Details.aspx?newsid=35 [https://perma.cc/U93P-EXDC]. In Section II.B, I discuss the IRS

Whistleblower program, and I raise the question of whether the IRS should encourage the

public to report tax-related social media posts to the agency.

SUPPLEMENTAL NUTRITION ASSISTANCE PROGRAM, supra note 58, at 23.

Id. at 17 (“Our testing found the recommended e-commerce monitoring tool less

effective than manual searches in detecting postings indicative of potential trafficking, and

we found the tool for monitoring social media to be impractical for states due to the volume

of irrelevant data.”).

Thomhave, supra note 63 (“[I]n the 1960s, welfare officials would regularly make

unannounced home visits (sometimes even ‘midnight raids’) to women receiving traditional

cash benefits to see if they lived in accordance with welfare eligibility rules.”). While not

referring specifically to social media mining, sociologist Spencer Headworth comments on

loss of privacy: “Making their lives transparent and legible to state agencies is one way the

poor pay for public assistance.” Headworth, supra note 36, at 172.

Thomhave, supra note 63 (emphasis in original).

See generally Michelle Lyon Drumbl, Beyond Polemics: Poverty, Taxes, and

Noncompliance, 14

EJOURNAL OF TAX RSCH. 253, 254 (2016) (identifying the high rate of

improper payments as one explanation of the high audit rate of EITC returns).

318 OREGON LAW REVIEW [Vol. 99, 301

2. Private Actor Use of Social Media Mining and Big Data

As noted, the use of artificial intelligence to screen individuals

extends beyond government agencies to the private sector, and some

companies are using social media algorithms as a screening tool, just

as the IRS seeks to do. One such online service company, Predictim,

ceased operation after it faced backlash for its methods.

Predictim

had used AI to scan social media posts to assess babysitters’

personalities.

This cyber-sleuthing service apparently appealed to

some anxious parents worried about leaving their child with an

unknown person; the company asserted that its algorithm could predict

such things as a sitter’s risk of drug use, tendency to bully, or even a

“bad attitude.”

Predictim provided parents a risk rating based on the

results of the algorithm, but the company did not explain how the

algorithm arrived at that risk rating, and the results were not shared

with the potential babysitter.

Though Predictim no longer offers this service, technology reporter

Drew Harwell observes that this type of technology is increasingly used

by companies in other types of hiring.

Harwell notes that companies

use such methods in recruiting, hiring, and reviewing their workers,

“offering employers an unrivaled look at job candidates through a new

wave of invasive psychological assessment and surveillance.”

Insurance companies are another class of private actor eager to

utilize information from social media, whether as a tool to determine

risk (and therefore premium rates) or to detect fraud.

One

Predictim halted its service after backlash from media publicity, including a

Washington Post article, and after it was blocked by Facebook, Twitter, and Instagram.

Drew Harwell, AI Start-Up That Scanned Babysitters Halts Launch Following Post Report,

ASH. POST (Dec. 14, 2018, 8:44 AM), https://www.washingtonpost.com/technology

/2018/12/14/ai-start-up-that-scanned-babysitters-halts-launch-following-post-report/ [https:

//perma.cc/TN2K-3EDD].

Drew Harwell, Wanted: The ‘Perfect Babysitter.’ Must Pass AI Scan for Respect

and Attitude, W

ASH. POST (Nov. 23, 2018, 8:50 AM), https://www.washingtonpost.com

/technology/2018/11/16/wanted-perfect-babysitter-must-pass-ai-scan-respect-attitude/

[https://perma.cc/WTB2-N7GQ] (“[Predictim uses] language-processing algorithms and an

image-recognition software known as ‘computer vision’ to assess babysitters’ Facebook,

Twitter and Instagram posts for clues about their offline life.”).

Id.

Id. (“[Tech firms sell to employers] artificial-intelligence systems that analyze a

person’s speech, facial expressions and online history with promises of revealing the hidden

aspects of their private lives.”).

Id.

Jessica Baron, Life Insurers Can Use Social Media Posts to Determine Premiums, as

Long as They Don’t Discriminate, FORBES (Feb. 4, 2019, 1:26 PM), https://www.forbes

2021] #Audited: Social Media and Tax Enforcement 319

commentator’s concern about insurance companies’ use of this practice

could be generalized as an overarching concern that any type of social

media mining can be potentially misleading.

“If you’ve given up

smoking but have old photos with cigarette in hand (or repost one of

those popular Facebook Memories) how can a computer (or even an

underwriter with a lot of work to do) properly assess the context of a

photo?”

Moreover, the image social media users cultivate online

often does not portray real life. In turn, this can limit the value of the

information obtained on social media and create time-consuming false

positives for the agency or company engaging in the search.

People seem willing to tolerate and consent to privacy losses coupled

with technology (often in connection with their smartphone) for a host

of reasons, ranging from convenience to reassurance to rewards.

Consider also the example of summer sleep-away camps offering a

facial-recognition service, with an app sending the camper’s parents a

notification the moment their child’s photograph is uploaded to its

site.

In 2019, the University of Alabama offered its students an

incentive to opt into cell phone–based location tracking because

football coach Nick Saban did not like that students were leaving the

stadium before games ended.

The school created an app that students

could download on their smartphone to verify that they attended and

stayed at games for the full four quarters of the contest; those who did

so accumulated reward points, which would be used to provide

preferred access to tickets for championship games.

During the first

.com/sites/jessicabaron/2019/02/04/life-insurers-can-use-social-media-posts-to-determine

-premiums/#7e18b19f23ce [https://perma.cc/SZW6-AFGW].

Id.

Id. In the context of life insurance, a cigarette represents a health risk for actuarial

purposes, as would certain hobbies such as skydiving or motorcycle riding. As other scholars

have noted, surveillance of poor people may focus on the use of cigarettes and alcohol, or

perhaps gambling. See E

UBANKS, supra note 30, at 114 (discussing the example of Maine

Governor Paul LePage).

See discussion infra Section II.A.

Drew Harwell, As Summer Camps Turn on Facial Recognition, Parents Demand:

More Smiles, Please, WASH. POST (Aug. 8, 2019, 1:26 PM), https://www.washingtonpost

.com/technology/2019/08/08/summer-camps-turn-facial-recognition-parents-demand-more

-smiles-please/?hpid=hp_hp-top-table-main_facecamp-815pm%3Ahomepage%2Fstory-ans

[https://perma.cc/UA47-7K3X].

Alex Scarborough, Bama Tracking Students to Check 4-Quarter Stays, ESPN

(Sept. 13, 2019), https://www.espn.com/college-football/story/_/id/27608647/bama-tracking

-students-check-4-quarter-stays [https://perma.cc/9NPQ-RYXZ].

Billy Witz, Orwellabama? Crimson Tide Track Locations to Keep Students at Games,

N.Y.

TIMES (Oct. 1, 2019), https://www.nytimes.com/2019/09/12/sports/alabama-tracking

-app.html [https://perma.cc/KM9X-AAS6] (“Greg Byrne, Alabama’s athletic director,

320 OREGON LAW REVIEW [Vol. 99, 301

game the incentive was available, so many students opted in that the

stadium’s network servers went down.

For better or for worse, these sorts of commercial uses of data

mining and surveillance may have the effect of desensitizing people,

especially younger generations, to the loss of privacy. But what is it

that we should actually expect of our revenue agency, or of our federal

government? The next section explores the taxpayer “right to privacy”

and examines these issues through a tax-specific lens.

B. What Is a Taxpayer’s Right to Privacy?

In 2015, Congress codified a Taxpayer Bill of Rights (TBOR),

which the Taxpayer Advocate refers to as a list of ten fundamental

rights that taxpayers should be aware of when dealing with the IRS.

Internal Revenue Code (the Code) § 7803(a)(3) provides, “In

discharging his duties, the Commissioner shall ensure that employees

of the Internal Revenue Service are familiar with and act in accord with

taxpayer rights as afforded by other provisions of this title,” and lists

the ten rights. Number seven on the list is “the right to privacy.”

There is a separately enumerated taxpayer right to confidentiality

(number eight on the list), which addresses how and with whom the

IRS can share taxpayer information.

By statutory implication, this

means the right to privacy is more than just a right to have one’s tax

information kept private from the public.

In his literature review of contemporary scholarship on tax and

privacy, Michael Hatfield categorizes the common characteristics that

frame the scholarly conversation about privacy.

With one notable

exception that relates to low-income taxpayers and the privacy

sacrifices they make in exchange for refundable tax credits,

Hatfield

said privacy concerns rarely came up when the program was being discussed with other

departments and student groups.”).

Id. Though not an Alabama fan or college student, my own teenage son told me he

would gladly consent to such tracking in order to secure College Football Playoff tickets.

I.R.C. § 7803; Taxpayer Bill of Rights, IRS (Nov. 25, 2020), https://www.irs.gov

/taxpayer-bill-of-rights [https://perma.cc/UH33-R82A].

I.R.C. § 7803(a)(3)(G). I.R.C. § 6103 provides a detailed set of limitations on how

and when the agency may disclose taxpayer information.

I.R.C. § 7803(a)(3)(H).

Hatfield, supra note 14, at 604–10.

See Hayes Holderness, Taxing Privacy, 21 GEO. J. POVERTY L. & POL’Y 1, 30–32

(2013) (expressing concern that the EITC, like other social welfare programs, imposes a

privacy burden on its recipients). As I discuss in the next section, this privacy burden

2021] #Audited: Social Media and Tax Enforcement 321

finds that “privacy” scholarship focuses on protection from the

disclosure of information (i.e., what the IRS can do with information it

collects) rather than how we conceive of privacy limitations on the

collection of information.

Hatfield describes the Code’s privacy

protections as “aim[ing] to maintain the confidentiality of taxpayer

information”

while noting that the IRS has authority “to collect any

information relevant to the 145,000,000 individual income tax returns

filed each year.”

If privacy and confidentiality are distinct rights and disclosure

rules primarily protect confidentiality, then what exactly is meant by

the “right to privacy?”

At the time of this writing, there is no case

law interpreting § 7803(a)(3)(G),

nor are there Treasury Regulations

providing formal administrative guidance on the various taxpayer

rights.

Adam Thimmesch, a tax scholar who has written about

intersects with the fact that low-income taxpayers are disproportionately selected for audit

when they claim the EITC.

Hatfield, supra note 14, at 606; see also Adam B. Thimmesch, Tax Privacy?, 90

TEMP. L. REV. 375, 375 (2018) (“The academic literature addressing privacy in the context

of the U.S. tax system has generally discussed tax privacy as nothing more than a limited

right of confidentiality.”).

Hatfield, supra note 14, at 596. Hatfield points to § 6103, which generally provides

that tax return information shall be kept confidential. Id. at 598. Hatfield notes that the IRS

can collect any relevant information about taxpayers without probable cause of a crime or

suspicion of a misstatement or understatement of any kind. Id. at 580.

Id. at 580 (describing examples of cases in which the IRS reviewed medical records,

love letters, family dynamics, reading habits, and other details of people’s private lives).

Hatfield argues that tax scholarship “has not addressed the risks of excessive information

collection.” Id. at 610.

See Joshua P. Law, Balancing Efficient IRS Administration and Taxpayer Rights, 43

ETON HALL LEGIS. J. 337, 348 (2019) (presenting a brief synopsis of the right to privacy

in a note that was published after Hatfield’s literature review) (“The [taxpayer’s] right to

privacy is ostensibly meant to guarantee taxpayers assurance that information about their

financial situation will not be intruded upon without due cause; however, the IRS currently

remains far from full compliance with this right.”).

Courts have stated that the Taxpayer Bill of Rights does not create any new rights. See

Moya v. Comm., 152 T.C. 182, 192 (2019) (“We think that the history of the IRS TBOR

makes clear that it accords taxpayers no rights they did not already possess.”); Facebook v.

IRS, No. 17-cv-06490-LB, 2018 WL 2215743, at *13 (N.D. Cal., May 14, 2018) (“The

statutory TBOR enacted as part of the 2015 PATH Act did not grant new enforceable

rights.”); Atl. Pac. Mgmt. Grp. v. Comm’r, 152 T.C. 330, 336 (2019) (“[S]ection 7803(a)(3)

itself does not confer any new rights on taxpayers; it merely lists ‘taxpayer rights as afforded

by other provisions of the Code.”).

Whether taxpayer rights are even enforceable is also an open question. See Alice G.

Abreu & Richard K. Greenstein, The U.S. Taxpayer Bill of Rights: Window Dressing or

Expression of Justice?, 4 J. TAX ADMIN. 25, 27 (2018); Leandra Lederman, Is the Taxpayer

Bill of Rights Enforceable? (Ind. Univ. Maurer Sch. of L., Working Paper No. 404, 2019),

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3365777 [https://perma.cc/WCE9

322 OREGON LAW REVIEW [Vol. 99, 301

privacy, refers to the TBOR’s statutory right to privacy as

“aspirational,” insofar as the standard is loose and subjective, and there

is no apparent remedy.

The IRS defines the right to privacy as follows: “Taxpayers have the

right to expect that any IRS inquiry, examination, or enforcement

action will comply with the law and be no more intrusive than

necessary, and will respect all due process rights, including search and

seizure protections, and will provide, where applicable, a collection due

process hearing.”

The IRS elaborates on the examination process as

follows:

The process of selecting a return for examination usually begins in

one of two ways. First, we use computer programs to identify returns

that may have incorrect amounts. These programs may be based on

information returns, such as Forms 1099 and W-2, on studies of past

examinations, or on certain issues identified by compliance projects.

Second, we use information from outside sources that indicates that

a return may have incorrect amounts. These sources may include

newspapers, public records, and individuals. If we determine that the

information is accurate and reliable, we may use it to select a return

for examination.

As a starting point, then, query whether the IRS looking at public

social media accounts or other publicly available online information

would constitute a violation of a taxpayer’s right to privacy. It seems

intuitive that if an individual makes information about him or herself

-3PJ7]. But see T. Keith Fogg, Can the Taxpayer Bill of Rights Assist Your Clients?, 91

TEMP. L. REV. 705, 729 (2019) (arguing that the taxpayer’s right to privacy has a role in

examination cases with respect to a revenue agent seeking information from third parties,

and in collection cases with respect to the loss of privacy that results from the filing of a

notice of federal tax lien).

Thimmesch, supra note 89, at 392–93. Thimmesch observes that tax design must

weigh privacy interests, whether those are described in neutral or normative terms, against

potential information-collection harms (among other potential harms). Id. at 414. Building

on the work of Daniel Solove, Thimmesch notes that “information-collection harms occur

regardless of whether the information is collected through surveillance or through direct

interrogation” and “regardless of whether the information is collected by the government or

by a private actor.” Id. at 412 (citing Daniel J. Solove, A Taxonomy of Privacy, 154 U.

PA.

L. REV. 477, 491–92 (2006)).

IRS, PUBL’N NO. 1, YOUR RIGHTS AS A TAXPAYER (Sept. 2017) (emphasis added),

https://www.irs.gov/pub/irs-pdf/p1.pdf [https://perma.cc/E7WQ-EG5C]. The right to

confidentiality is described distinctly: “Taxpayers have the right to expect that any

information they provide to the IRS will not be disclosed unless authorized by the taxpayer

or by law.” Id.

Id.

2021] #Audited: Social Media and Tax Enforcement 323

publicly available on such accounts, that individual has no reasonable

expectation of privacy with respect to what has been posted.

To reiterate, the primary concern of this Article is not whether the

IRS looking at social media violates the taxpayer’s right to privacy or

whether the IRS has the right to access and use this information for

enforcement purposes. I think the better question is whether, as a matter

of social policy, it is appropriate and desirable, or even cost-effective,

for the IRS to mine social media for civil enforcement purposes. As the

prior section explained, other government agencies can and do access

social media to investigate various types of benefits fraud. But most tax

noncompliance does not constitute “fraud.”

1. IRS Use of Data Analytics—Past and Future

The IRS has been using data analytics to address noncompliance

since long before Facebook and Instagram. Since the 1960s, the agency

has used something called Discriminant Function (DIF) scoring to

select returns for audit.

The DIF score is a three-digit score assigned

by a series of algorithms and models, the contents of which are not

known outside the IRS.

100

According to the IRS, the DIF “score rates

the potential for [tax liability] change, based on past IRS experience

with similar returns”—with a higher score more likely to be selected

for examination.

101

The DIF is used as a screening tool, and a return

identified by a DIF score is manually classified by an “experienced

[IRS] examiner” in order to screen for significant issues “worthy of

exam.”

102

The examination selection process has evolved over time through

the use of technology and automation, and over time these methods

have increased the efficiency of audits by using data to identify returns

Stacey Vanek Smith, When the IRS ‘Likes’ Your Facebook Update, MARKETPLACE

(Apr. 14, 2014) https://www.marketplace.org/2014/04/14/when-irs-likes-your-facebook

-update/ [https://perma.cc/KJQ9-UT38]. As tax scholar Edward Zelinsky remarked, “It’s

hard to believe that anybody who puts anything on Facebook has any legitimate expectati

of privacy.” Id.

See Carina Federico & Travis Thompson, Do IRS Computers Dream About Tax

Cheats? Artificial Intelligence and Big Data in Tax Enforcement and Compliance, J. TAX

PRAC. & PROC., Feb.–Mar. 2019, at 43, 45; TOM GREENAWAY & ALEXANDRA DESANTIS,

Taking the Mystery Out of Examinations—the Audit Process, in 1 E

FFECTIVELY

REPRESENTING YOUR CLIENT BEFORE THE IRS 3-3 (Keith Fogg ed., 7th ed. 2018).

100

GREENAWAY & DESANTIS, supra note 99, at 5; see also IRM 4.1.5.3.3.1(1) (Sept.

21, 2020).

101

I.R.S. Fact Sheet FS-2006-10 (Jan. 2006).

102

IRM 4.1.5.3.3.1 (Sept. 21, 2020) (“Classifiers must use their skills, technical

expertise, local knowledge, and experience to identify hidden, as well as obvious, issues.”).

324 OREGON LAW REVIEW [Vol. 99, 301

more likely to be noncompliant.

103

Other examples include the IRS

automated underreporter program that matches returns with third-party

reporting and the requirement implemented in 1987 to list identification

numbers of dependents.

104

The IRS also screens individual income tax

returns using databases external to the agency to detect errors and

noncompliance. For example, returns claiming the EITC are screened

against the Dependent Database, which includes information from

external sources.

105

These returns are scored according to certain

indicators, with a higher score suggesting a greater probability that a

taxpayer may not meet the credit’s residency, relationship, or age

eligibility requirements.

106

While the public does not know how DIF scores are calculated, or

exactly what information is entered into the Dependent Database and

from which sources, it stands to reason that the possibilities of social

media data mining extend far beyond the traditional tools used for

return screening.

In their 2017 article, Kimberly Houser and Debra Sanders detail how

the IRS is already engaging in data mining of public and commercial

data pools, including social media sites such as Facebook, Instagram,

and Twitter, and using that information to run predictive algorithms.

107

In addition to raising procedural and due process concerns about this

practice,

108

Houser and Sanders also express a concern that the IRS’s

use of such algorithms may result in discrimination.

109

103

See generally Houser & Sanders, supra note 14, at 828–33.

104

Id. at 829; see also James Alm et al., New Technologies and the Evolution of Tax

Compliance, 39 VA. TAX REV. 287, 312–13.

105

See NAT’L TAXPAYER ADVOC., THE EARNED INCOME CREDIT, 3 SPECIAL REPORT

CONGRESS 2020, at 5 n.26. (“The [Dependent Database] is a rule-based system

incorporating data within the IRS and information from external sources such as the

Department of Health and Human Services and the Social Security Administration.”); see

also IRM 4.19.14.1.1 (Dec. 7, 2017) (“Exam receives the majority of its EITC work from

the Dependent Database [] and Electronic Fraud Detection System.”).

106

NAT’L TAXPAYER ADVOC., supra note 105, at 27. The Dependent Database is

referenced in IRM 4.19.14.8 (Apr. 4, 2019), but most of the details are redacted.

107

Houser & Sanders, supra note 14, at 819–20. Houser and Sanders also detail other

types of privacy breaches not discussed in this Article, including how the ACLU has found

that the IRS purchased cell phone tracking technology and also discovered in 2013 that the

IRS policy permitted the reading of private emails without a warrant (it has since agreed to

stop doing so). Id. at 822–23.

108

Id. at 843 (identifying ways in which the IRS data collection and analytics practices

may violate the Administrative Procedures Act and/or due process protections).

109

Id. at 848–50 (“[Algorithms] may result in targeting certain groups based on the

associations created as the algorithm learns . . . . [And t]he New York Police Department

came under fire for its use of predictive analytics to focus its policing on certain

communities.”). But see generally I. Bennett Capers, Race, Policing, and Technology,

2021] #Audited: Social Media and Tax Enforcement 325

My greatest concern, which Houser and Sanders do not specifically

raise, is that data mining of social media sources will disproportionately

impact low-income taxpayers. As I

110

and others

111

have highlighted

elsewhere, low-income taxpayers are already disproportionately in

the examination spotlight: in fiscal year 2018, the audit rate for all

individual taxpayers was 0.59%; when broken down by level of

adjusted gross income (AGI), a higher percentage of taxpayers with

AGI between $1 and $25,000 were audited (0.69%) than those with

AGI between $25,000 and $499,000.

112

As the next section addresses,

audits of low-income taxpayers are typically time-consuming,

burdensome, and stressful. Taxpayers are asked to provide

documentation substantiating fact-intensive questions, such as who

lives in their household, what their relationship to those individuals is,

and how much support the taxpayer provides to those individuals.

113

For low-income taxpayers, audits can pose an economic hardship for

those cases in which the IRS freezes the valuable refundable credit

portion of the refund until the audit is resolved.

114

95 N.C. L. REV. 1241 (2017) (advocating for the harnessing of surveillance technology to

deracialize policing and describing the necessary redistribution of privacy).

110

See MICHELLE LYON DRUMBL, TAX CREDITS FOR THE WORKING POOR (2019).

111

Kiel, supra note 19 (“[A]udits of the rich continue to plunge while those of the poor

hold steady, and the two audit rates are converging . . . . [EITC recipients] are audited at a

higher rate than all but the richest taxpayers.”).

112

I.R.S., PUB. NO. 55B, DATA BOOK: 2018, at 27 (May 2019). At AGI levels above

$500,000, the audit rate is higher, and the tiny number of the taxpayers at the highest end

are subject to a sharply higher rate: the audit rate of taxpayers with AGI between $500,000

and $1 million was 1.1%; taxpayers with AGI between $1 million and $5 million had an

audit rate of 2.21%; taxpayers with AGI between $5 million and $10 million faced an audit

rate of 4.21%, and 6.66% of taxpayers with AGI above $10 million were audited. Id. Note

that taxpayers who reported no AGI were also subject to a higher than average audit rate

(2.04%); this includes returns with losses, and a taxpayer with no AGI cannot claim EITC.

Id.

113

See I.R.C. §§ 2(b), 7703(b), 24, 32, 151 and accompanying Treasury Regulations

(setting forth the eligibility requirements for head of household filing status, marital status,

the child tax credit, the EITC, and the dependent exemption). These various eligibility

requirements overlap imperfectly and can be particularly complicated to apply to

multigenerational households and shared child custody situations. Examiners ask taxpayers

to prove residence of individuals through the use of such documents as school records,

medical or social service records, and court orders, if applicable. See, e.g., I.R.S. Form 886-

H-EIC (Oct. 2019), https://www.irs.gov/pub/irs-pdf/f886he.pdf [https://perma.cc/Y2LN-

YAKZ].

114

See infra text accompanying note 120.

326 OREGON LAW REVIEW [Vol. 99, 301

2. Disproportionate Enforcement on Lower-Income Taxpayers, and

Implications for Taxpayer Privacy

The higher audit rate of these low-income taxpayers, relative to the

moderate-income taxpayers, is due in part to the fact that the IRS

prioritizes audits of returns claiming the EITC. In recent years, more

than one in three individual income tax returns selected for examination

involved an EITC claim.

115

The EITC remains an enforcement priority

for the IRS even as other types of examinations have declined and even

though the overall number of returns with EITC claims has not

increased in recent years.

116

The IRS has been subject to criticism for

shining the spotlight on the poor rather than the rich, as well as for the

demographic consequences of this enforcement strategy.

117

A former

IRS economist who analyzed exam coverage data and estimated how it

breaks down by county and state reported that taxpayers in the rural

south are being audited at disproportionately high rates; according to

his estimates, the ten most-heavily audited counties in the United States

are disproportionately rural, low-income, and nonwhite.

118

115

In 2018, 37% of audited returns were selected on the basis of an EITC claim. I.R.S.,

PUB. NO. 55B, supra note 112, at 23–26. Table 9a shows that 150,043,227 individual

income tax returns were filed; of this total, 892,187 were selected for audit, representing an

overall individual income tax audit rate of 0.59% for FY 2018. Id. Of the 892,187 returns

selected for audit, footnote 5 specifies that 330,359 (37%) were selected for audit on the

basis of an EITC claim. Id. at 126 n.5. See also Paul Kiel & Jesse Eisinger, Who’s

More Likely to Be Audited: A Person Making $20,000—or $400,000?, P

ROPUBLICA

(Dec. 12, 2018, 5:00 AM), http://www.propublica.org/article/earned-income-tax-credit-irs-

audit-working-poor [https://perma.cc/HMP7-DDVR] (describing how EITC audits have

increased as a proportion of audits over a period of time in which IRS resources have been

reduced) (“[In FY 2016], the IRS audited 381,000 recipients of the EITC. That was 36

percent of all audits the IRS conducted, up from 33 percent in 2011, when the budget cuts

began.”).

116

See, e.g., TREASURY INSPECTOR GEN. FOR TAX ADMIN., NO. 2019-30-063, TRENDS

COMPLIANCE ACTIVITIES THROUGH FISCAL YEAR 2018, at 15 (Sept. 9, 2019) (“Overall,

the number of tax returns claiming the [EITC] have not increased over the past five years.

At the same time, the number of examinations conducted by the [Wage & Investment]

Division for returns claiming the [EITC] have increased by 17 percent, from 282,665 in FY

2014 to 330,886 in FY 2018.”); see also George testimony, supra note 16, at 10 (noting that

the IRS added 290 additional tax examiner positions, that tax examiners were the only

position the IRS increased in 2018 and that correspondence examinations were the only

types of examinations that the IRS increased in 2018. In contrast, the IRS has decreased the

number of revenue agent positions; revenue agents conduct more complex examinations in

the field).

117

See, e.g., Paul Kiel & Hannah Fresques, Where in the U.S. Are You Most Likely

to Be Audited by the IRS?, PROPUBLICA (Apr. 1, 2019), https://projects.propublica.org

/graphics/eitc-audit [https://perma.cc/JZ7F-EJLL].

118

Kim M. Bloomquist, Regional Bias in IRS Audit Selection, 162 TAX NOTES 987

(Mar. 4, 2019) (discussing regional bias in audits for tax years 2012–2015 and highlighting

2021] #Audited: Social Media and Tax Enforcement 327

The majority of all examinations, and EITC examinations in

particular, are correspondence examinations rather than in-person

audits.

119

Correspondence examinations, as the name suggests, are

conducted through the mail. These examinations are highly automated

and are not overseen by one specific employee from start to finish. In

other words, the IRS uses an inefficient process to audit low-income

taxpayers on an issue with fact-intensive and complex eligibility

requirements. These audits are time-consuming to resolve, with the

result that refunds of thousands of dollars are often delayed for

months or longer.

120

Further, these audits are often burdensome for

taxpayers.

121

Yet the IRS has justified this disproportionate focus on

EITC claimants, citing resource constraints.

122

At the same time, the

IRS has faced criticism for not doing more to pursue high-income

individuals who fail to file a tax return, even though pursuing nonfilers

is also cost-efficient because of automation and the use of third-party

information reporting.

123

that rural Humphreys County, Mississippi, with a median annual household income of just

$26,000, had the highest rate of audit intensity, while also noting that the ten most heavily

audited counties are all in the rural south and that in 2017 the population of these ten counties

was 79% nonwhite (primarily African-American) according to the U.S. Census Bureau).

119

Seventy-one percent of examinations conducted in FY 2017 were correspondence

examinations, which is fairly typical. NAT’L TAXPAYER ADVOC., ANNUAL REPORT TO

CONGRESS 126 (2018), http://www.TaxpayerAdvocate.irs.gov/2018AnnualReport [https://

perma.cc/PWB6-Z3LM]. In the same year, approximately 72% of correspondence exams in

the Wage & Income Division were EITC exams. Id. at 128.

120

Most EITC exams occur prior to the issuance of the refund, with the result that the

refund is frozen and the taxpayer does not receive the EITC until the audit is concluded in

the taxpayer’s favor. Id. at 130, fig.1.8.3. Depending on household composition and income

level, these refunds are often thousands of dollars, making them significant relative to the

annual earned income of the household; low-income households rely on these refunds as a

critical anti-poverty safety net. See C

ONG. RSCH. SERV. REP., THE EARNED INCOME TAX

CREDIT (EITC): HOW IT WORKS AND WHO RECEIVES IT, R43805 (Jan. 12, 2021),

https://fas.org/sgp/crs/misc/R43805.pdf [https://perma.cc/33EY-LTPX].

121

Id. at 133 (citing lower education levels, reliance on tax return preparers, and

language barriers as some of the challenges faced by EITC claimants upon audit).

122

Describing how the IRS audits approximately 300,000 EITC returns a year, the

agency’s website states, “EITC correspondence audits are the most efficient use of available

IRS examination resources with the average time to complete the audit of 5 hours per

return.” I.R.S. Update on Audits (Sept. 24, 2020), https://www.irs.gov/newsroom/irs-update

-on-audits [https://perma.cc/WL6P-VLB4].

123

TREASURY INSPECTOR GEN. FOR TAX ADMIN., NO. 2020-30-015, HIGH-INCOME

NONFILERS OWING BILLIONS OF DOLLARS ARE NOT BEING WORKED BY THE INTERNAL

REVENUE SERVICE (May 29, 2020) [hereinafter TREASURY INSPECTOR REPORT]. As the

TIGTA report notes, the intentional failure to file federal income tax returns is a crime.

Id. at 1. But see I.R.S. News Release IR-2020-34 (Feb. 19, 2020) (announcing that revenue

officers would increase in-person visits to taxpayers with income of more than $100,000

who did not file returns in 2018 or previous years). IRS Commissioner Charles Rettig

328 OREGON LAW REVIEW [Vol. 99, 301

It should be noted that while EITC claimants are disproportionately

in the IRS enforcement spotlight, recipients of this social benefit are in

many respects perceived to be treated with more dignity than recipients

of traditional welfare.

124

Tax scholar Hayes Holderness describes the

EITC as “far less privacy-invasive than either TANF or SNAP,” as both

of those social benefit programs require precertification whereas the

EITC does not.

125

Moreover, as Zelenak describes, it is rare for the IRS

to make a criminal referral for an individual EITC claimant; individuals

who wrongfully claim the EITC usually face only civil sanctions.

126

3. Differing Policy Implications for the IRS Examination and

Collection Functions

The IRS RFI explicitly mentions the agency’s interest in using social

media research tools for “existing compliance cases” and mentions

collections as an end goal.

127

Though it does not explicitly mention the

potential use in examinations, it is not implausible that the IRS would

consider using such tools to access social media in the context of both

examinations and collections.

While social media is a relatively new phenomenon, the IRS has

long navigated criticism and limitations on the appropriate balance

between taxpayer privacy and tax enforcement, including in

examinations. In the early 1990s, the IRS acknowledged it had

expanded the use of its “economic reality” investigations; whereas the

indirect method technique was previously used only in criminal cases

involving unfiled returns, the IRS extended its use of these so-called

emphasized this initiative as a priority in testimony before the Senate Finance Committee

in June 2020: “We continue working towards the goal of having a presence in every

neighborhood, on each type of tax issue and at every level of income, to ensure fairness

for all taxpayers.” 2020 Filing Season and COVID-19 Recovery: Before the S. Fin.

Comm., 116th Cong. (June 30, 2020) (Testimony of Charles P. Rettig, Commissioner,

Internal Revenue Service), https://www.finance.senate.gov/hearings/2020-filing-season-

and-irs-covid-19-recovery [https://perma.cc/T6GR-VS8G].

124

See generally SARAH HALPERN-MEEKIN ET AL., IT’S NOT LIKE I’M POOR (2015)

(recounting interviews with EITC recipients who speak to their perceptions); see also Sara

Sternberg Greene, The Broken Safety Net: A Study of Earned Income Tax Credit Recipients

and a Proposal for Repair, 88 N.Y.U. L. REV. 515 (2013).

125

Hayes Holderness, Taxing Privacy, 21 GEO. J. ON POVERTY L. & POL’Y 1, 30 (2013).

126

To the extent that the IRS makes criminal referrals in connection with the EITC, it is

most commonly in the context of tax return preparer fraud and promoters of fraudulent

schemes, rather than individual claimants. Zelenak, supra note 56, at 1891–92. One might

analogize return preparer fraud to social benefit trafficking.

127

IRS RFI, supra note 2.

2021] #Audited: Social Media and Tax Enforcement 329

lifestyle audits to civil cases.

128

Agents were trained to look for

economic signs that taxpayers were living beyond their means,

signaling possible unreported income.

129

Tax professionals

complained that the audits were arbitrary, blurred the lines between

civil examinations and criminal investigations, and involved intrusions

into taxpayer privacy.

130

Congress subsequently enacted § 7602(e),

which limited the circumstances under which the IRS could use this

technique.

131

While § 7602(e) seemingly gave taxpayers a new defense

and even an opportunity to invoke a due process right, the statute

provides only that the IRS must show it had a “reasonable indication”

of a “likelihood” of unreported income. The relatively thin caselaw on

§ 7602(e) suggests that the IRS has had little trouble in meeting this

standard to justify its use of lifestyle audits when challenged by

taxpayers.

132

In its internal legal advice interpreting § 7602(e), the

Office of Chief Counsel advised IRS revenue agents that they may

drive by a taxpayer’s house, or may conduct a Lexis search to ascertai

if the taxpayer purchased real estate in a year at issue, prior to having a

reasonable indication that there is a likelihood of unreported income.

133

All this raises the question: If the IRS decides to pursue the use of

social media in both examinations and collections, should it develop

different policies and guidelines for the use of social media by these

distinct functions?

128

Albert B. Crenshaw, Tax Cheats Beware: The IRS Will Now Audit Lifestyles, WASH.

POST (Nov. 6, 1994), https://www.washingtonpost.com/archive/business/1994/11/06/tax

-cheats-beware-the-irs-will-now-audit-lifestyles/1c220596-4ee3-4ec3-80cb-b06a0e443c32/

[https://perma.cc/T6GR-VS8G].

129

See, e.g., Philip R. Fink & Charles Gibson, Less Reason to Be Afraid?, CPA J. (June

1999), http://archives.cpajournal.com/1999/0699/features/f46699.html [https://perma.cc

/6NC9-UU2E].

130

Id.; Barbara Whitaker, Spending It; When the I.R.S. Agent Peeks Under the Mattress,

N.Y. T

IMES (July 28, 1996), https://www.nytimes.com/1996/07/28/business/spending-it

-when-the-irs-agent-peeks-under-the-mattress.html [https://perma.cc/J648-GH74].

131

I.R.C. § 7602(e) (“The Secretary shall not use financial status or economic reality

examination techniques to determine the existence of unreported income of any taxpayer

unless the Secretary has a reasonable indication that there is a likelihood of such unreporte

income.”).

132

See, e.g., Hsu v. United States, No. 17-cv-06656 NC, 2018 WL 2234439 (N.D. Cal.

May 16, 2018); Mortland v. IRS, No. A-03-CA-115-SS, 2003 WL 21791249 (W.D. Tex.

June 24, 2003); United States v. Abramson-Schmeiler, No. 09-cr-00359-REB, 2010 WL

11537887 (D. Colo. Oct. 14, 2010); Chapin v. IRS Agent, No. 2:14-cv-538-EJL-REB, 2016

WL 383135 (D. Idaho Jan. 8, 2016).

133

I.R.S. Assoc. Chief Couns. Mem. 200101030 (Oct. 25, 2000) (concluding that these

specific practices would not constitute “an intrusion on a taxpayer”).

330 OREGON LAW REVIEW [Vol. 99, 301

The IRS examination function and the IRS collection function each

play a role in enforcement. As there are slight distinctions in their

mission, each function’s use of social media mining and big data

algorithms might have different implications.

134

IRS policy states that,

in selecting returns for examination,

[t]he primary objective . . . is to promote the highest degree of

voluntary compliance on the part of taxpayers. This requires t

exercise of professional judgment in selecting sufficient returns of all

classes of returns in order to assure all taxpayers of equitable

consideration, in utilizing available experience and statistics

indicating the probability of substantial error, and in making the mos

efficient use of examination staffing and other resources.

135

The IRS has a separately stated set of policy considerations for

collections, which also emphasizes that “enforcement is a necessary

component of a voluntary assessment system.”

136

Unlike in the

examination context, taxpayers have additional rights and protections

in the collections context, some of which are statutory

137

and some of

which are a function of IRS policy.

138

For example—and of critical

importance to low-income individuals—the IRS cannot pursue forced

collections through levy if doing so will create an economic hardship

for the taxpayer.

139

In addition to statutory notice and due process

requirements, there are additional checks and balances in the collection

process. For example, by statute, the IRS cannot seize the taxpayer’s

principal residence without the approval of a U.S. district court judge

or magistrate.

140

134

Recall that the RFI references using social media mining tools only with respect to

collections and explicitly mentions that it would limit the use to open cases. See supra note

2. I include both examination and collections in this discussion in light of Houser a

Sanders’ concerns about how the IRS is using its big data database. See Houser & Sanders,

supra note 14, at 870.

135

IRM 1.2.1.5.10 (June 1, 1974).

136

IRM 1.2.1.6.1 (August 18, 1994).

137

See I.R.C. §§ 6300–44, which includes notice and demand requirements and

collection due process rights related to the imposition of liens and levies.

138

See, e.g., IRM 5.11.7.2.1.1(2)(e) (Sept. 23, 2016); I.R.S., Dir. of Collection Inventory

Delivery and Selection Mem. SBSE-05-1015-0067 (Oct. 7, 2015) (notifying IRS Field

Collection agents of a policy decision to exclude social security disability insurance

payments from the automated levy program, even though § 6331(h) provides the IRS the

statutory authority to levy those benefits).

139

I.R.C. § 6343(a); Treas. Reg. § 301.6343-1(b)(4) (as amended in 2005) (defining

economic hardship as a taxpayer being unable to pay reasonable basic living expenses).

140

I.R.C. § 6334(e). Section 6334(a)(13) additionally provides that any real property

used as the taxpayer’s residence is exempt if the amount of the levy does not exceed $5,000.

See also Treas. Reg. § 301.6334-1(d).

2021] #Audited: Social Media and Tax Enforcement 331

The IRS Policy Statement on “Fairness and Integrity in Enforcement

and Collection” states three primary goals: (1) to ensure fairness to the

taxpaying public; (2) to ensure an equitable process for all taxpayers;

and (3) to ensure fairness to each taxpayer.

141

The policy further

describes how it aims to ensure an equitable process: “[F]airness and

integrity are built into the foundation of our enforcement selection

processes. These processes operate under a comprehensive set of

checks and balances and safeguards to identify the highest potential

noncompliance, using scoring mechanisms, data driven algorithms,

third party information, whistleblowers and information provided by

the taxpayer.”

142

We should expect the IRS to pursue collection where

necessary to promote compliance, but it must take measures to avoid

undue hardship to the taxpayer.

143

As one way of balancing these enforcement goals with taxpayer

rights, Section II.C proposes that the IRS clarify its conception of the

taxpayer right to privacy to the public. The agency should be

transparent to the public about its use (if any) of social media, so that

taxpayers are on notice that the IRS is watching. It is hard, if not

impossible, to predict what compliance impact would result from

taxpayers knowing that the IRS is monitoring their social media.

144

However, putting taxpayers on notice of this intention affords them

dignity.

145

141

I.R.S. Policy Statement 1-236 (Oct. 24, 2016).

142

Id.

143

See, e.g., IRM 1.2.1.6.5 (July 10, 1959); IRM 1.2.1.6.14 (Nov. 19, 1980).

144

There is a rich literature on theories of tax compliance. See, e.g., Marjorie E.

Kornhauser, Normative and Cognitive Aspects of Tax Compliance: Literature Review and

Recommendations for the IRS Regarding Individual Taxpayers, in

NAT’L TAXPAYER

ADVOC., 2 ANN. REP. TO CONG. 138 (2007). One theory of tax compliance is economic

deterrence, which is the idea that taxpayers who perceive that the IRS is likely to catch tax

cheats will be deterred from cheating. However, studies suggest that deterrence is not the

strongest driver of tax compliance: other compliance motivations include social norms, tax

morale, trust in government, convenience, and reliance on tax preparers. Id.

145

Arguably, it also affords taxpayers who engage in noncompliance a warning to

cultivate their social media in a way that does not publicly tip off the IRS while continuing

to engage in such noncompliance. On the other hand, if everyone on social media cultivates

their online image with this in mind, with the effect that no one on social media brags about

cheating on their taxes, might this boost tax morale? These questions are worthy of research,

but beyond the scope of this Article.

332 OREGON LAW REVIEW [Vol. 99, 301

4. Differing Policy Implications for Civil Tax Enforcement and

Criminal Tax Enforcement

Just as there ought to be different considerations for examination and

collections, there should be different considerations in place for civil

proceedings and criminal investigations. As a policy matter, it may be

more justifiable for the IRS to engage in social media mining (along

with other use of artificial intelligence) to determine whether a criminal

act has been committed.

146

The IRS Criminal Investigations unit has

far fewer employees than the civil enforcement side,

147

and its smaller

workforce has a broader scope that includes not just tax fraud but also

money laundering and narcotics-related and counterterrorism-related

financial crimes.

148

Don Fort, chief of the IRS Criminal Investigations

unit, has spoken publicly about how his office is using “advanced data

analytics to spot suspicious behavior,” specifically mentioning his

office was using data analytics to investigate noncompliance with

payroll tax laws.

149

The need for this stems in part from the agency’s

decrease in staffing.

150

Fort has also made public remarks about

unreported capital gains related to cryptocurrencies as an emerging area

of focus and said that there are whistleblower cases they are actively

working; he cited cryptocurrencies as an area of concern because of the

lack of transparency and visibility in those transactions.

151

146

See generally Alm et al., supra note 104 (discussing ways that technology may both

improve and subvert tax compliance, including in the criminal tax context).

147

In fiscal year 2018, examinations and collections had 30,876 full-time equivalent

positions. IRS,

DATA BOOK 2018 at 67, tbl.30 (2018). In contrast, there were 2,019 special

agent positions in the Criminal Investigations unit in fiscal year 2018. IRS, IRS:

CRIMINAL INVESTIGATION ANNUAL REPORT 7 (2018), https://www.irs.gov/pub/irs-utl

/2018_irs_criminal_investigation_annual_report.pdf [https://perma.cc/GR3W-K5PA].

148

IRS, IRS: CRIMINAL INVESTIGATION ANNUAL REPORT 6 (2018).

149

Kristin Broughton, Tax Crime Enforcement Unit Relying More on Analytics to Spot

Crime, WALL ST. J. (June 12, 2019, 5:14 PM), https://www.wsj.com/articles/tax-crime

-enforcement-unit-relying-more-on-analytics-to-spot-crime-11560363826 [https://perma.cc

/JC7V-KHTW].

150

Id. (noting that the IRS Inspector General reports the IRS has 26% fewer special

agents than it did in 2012); see also Michael Cohn, IRS Criminal Investigation Leveraging

More Data Analytics, A

CCT. TODAY (Nov. 14, 2018, 12:29 PM), https://www.accounting

today.com/news/irs-criminal-investigation-leveraging-more-data-analytics-to-probe-tax

-fraud [https://perma.cc/PW5S-B6VU] (“One reason why the division needed to rely so

much on data analytics was to make up for the shortage of special agents, while helping IRS

agents identify the cases that would have the most impact.”).

151

Broughton, supra note 149. The IRS Criminal Investigation unit is devoting

significant resources to cryptocurrency investigations. See Federico & Thompson, supra

note 99.

2021] #Audited: Social Media and Tax Enforcement 333

Perhaps it is inevitable that the IRS will increasingly utilize artificial

intelligence techniques in criminal investigations, and perhaps the

public at large will not find the use of such techniques objectionable in

this context. If we embrace this as acceptable, where might this

ultimately take us? Will the IRS use facial recognition techniques to

track where taxpayers are going and to monitor their cash purchases?

152

Where, as a society, do we want this to end? And do we want this

technology to be applied to all taxpayers and to all potential tax-related

issues, both big and small?

153

UNISHING THE UNSOPHISTICATED: PONDERING BRAGGADOCIO,

WHISTLEBLOWERS, AND THE QUEST TO CLOSE THE TAX GAP

This Article is not meant to deny or undermine the importance of tax

compliance or IRS enforcement efforts. The IRS estimates there is a

relatively high and steady rate of voluntary tax compliance, most

recently estimated at approximately 83.6%.

154

The term “tax gap”

refers to the amount of tax not paid voluntarily and timely and refers to

three distinct types of noncompliance: nonfiling, underreporting, and

underpayment.

155

The IRS uses noncompliance estimates and research

programs to inform its enforcement strategies. For example, there is a

well-established correlation between information reporting and

compliance, and there is a well-known compliance gap correlating with

the cash economy due to the lack of information reporting. Underlying

the goal of reducing the tax gap (as well as underlying the RFI) are

152

EUBANKS, supra note 30, at 7 (citing the example of Maine Governor Paul LePage

having his administration mine data and publicly release information about welfare

recipients who withdrew cash from ATMs in smoke shops, liquor stores, and out-of-state

locations).

153

According to TIGTA, “[i]n the past, the Internal Revenue Service (IRS) has focused

on the tax compliance of high-income individuals because their noncompliance can have a

significant corrosive effect on tax administration.” T

REASURY INSPECTOR REPORT, supra

note 123, at 1. This is not to argue that the IRS should ignore small indiscretions, but given

the agency’s high focus on EITC enforcement, it ought to use technology to increase its

efforts at reducing high-dollar noncompliance. To see the proposal I set forth for this, see

my discussion in Section II.C.5.

154

IRS, PUBLICATION 1415, FEDERAL TAX COMPLIANCE RESEARCH: TAX GAP

ESTIMATES FOR TAX YEARS 2011–2013, at 1 (2019). This figure is the voluntary

compliance rate, which is a measurement of the gross tax gap; the net compliance rate, which

includes amounts that are eventually paid, including through enforcement and collection

measures, is a slightly higher figure (85.8% in latest report).

155

Id. at 4.

334 OREGON LAW REVIEW [Vol. 99, 301

questions about how to most efficiently allocate limited resources in

maximizing tax compliance.

For some, it may be easy to shrug off a call to protect the privacy of

people who voluntarily make public the intimate (and often mundane)

details of their private life. Section II.A addresses that temptation and

raises examples of when and why it might be unfair, unwise, or

inefficient to incorporate social media mining for tax enforcement. A

distinction can and should be made between blatant public admissions

of fraudulent behavior, on the one hand, and subtleties perceived by

possibly discriminatory algorithms that may lead to significant false

positives. As to blatant admissions, Section II.B explores the question

of whistleblowers and whether the IRS should expressly encourage

third parties to report potential fraud based on social media postings.

This part then concludes in Section II.C with several suggested

policy proposals for the IRS, including a call for the IRS to clarify its

understanding of a taxpayer’s “right to privacy” for the public.

A. The Temptation to Disregard Individual Privacy Concerns

One response to this Article can be couched in the words of the

fictional character Forrest Gump: “Momma says stupid is as stupid

does.”

156

In other words, people may intuitively respond to the IRS RFI

with a shoulder shrug: if taxpayers are foolish enough to get caught

doing something wrong, why is that necessarily bad? And if they are

bragging about illegal behavior publicly on the internet, why do they

deserve sympathy or leniency?

157

As one unsympathetic example of internet braggadocio, consider

John McAfee, the British American businessman who founded the

McAfee Associates software company. McAfee, a self-described

Libertarian who unsuccessfully ran for President in 2016, is a

multimillionaire who has publicly stated that he is a nonfiler.

158

I would

be surprised if this fact had been previously unknown to the IRS—

presumably at least some of his income is subject to third-party

156

FORREST GUMP (Paramount Pictures 1994).

157

Note that this dovetails with one of the major critiques of privacy rights: that “those

with nothing to hide have no reason to be concerned for privacy.” Hatfield, supra note 14,

at 591 (evoking arguments made by Judge Richard Posner and Catherine MacKinnon).

Several people I have spoken to about this Article had that reaction as well.

158

Brooker Crothers, John McAfee Is Running from U.S. Authorities – and Running for

President. On a Boat., FOX NEWS (Jan. 24, 2019), https://www.foxnews.com/tech/john

-mcafee-is-running-from-u-s-authorities-and-running-for-president-on-a-boat [https://perma

.cc/WQ7D-LGQZ].

2021] #Audited: Social Media and Tax Enforcement 335

information reporting,

159

and the IRS would easily determine if no

return had been filed in a given year. Of course, it is the height of hubris

to brag about this publicly on Twitter, as McAfee did:

160

I have not filed a tax return for 8 years. Why? 1: taxation is illegal.

2: I paid tens of millions already and received Jack Shit in services.

3. I’m done making money. I live off of cash from McAfee Inc. My

net income is negative. But i [sic] am a prime target for the IRS. Her

I am.

According to McAfee, he has told the IRS, “I am not filing a return,

I have no intention of doing so, come and find me.”

161

Though the IRS

has not publicly confirmed this, McAfee claims that in January 2019

the IRS convened a grand jury to charge him, his wife, and four

campaign workers “with unspecified IRS crimes.”

162

Another unsympathetic example of such public bluster is Rashia

Wilson, the self-declared “First Lady” and “Queen of Tax Fraud.”

163

Wilson pleaded guilty to one count of wire fraud and one count of

aggravated identity theft in connection with charges that she filed false

federal income tax returns using fraudulently obtained social security

numbers.

164

According to news reports, Wilson “taunted police by

posting photos of herself flashing stacks of cash”

165

and posted

inflammatory statements online, including: “I’M RASHIA, THE

QUEEN OF IRS TAX FRAUD . . . I’m a millionaire for the record, so

if U think indicting me will B easy it won’t, I promise you!”

166

159

The Internal Revenue Code imposes an information reporting requirement on certain

third-party payors. See I.R.C. § 6031. The list is voluminous and includes payors of rents,

royalties, wages, and sales or redemptions of securities. IRS research analysts have

consistently found a relationship between third-party information reporting and tax

compliance. See IRS,

PUBLICATION 1415, supra note 154, at 13.

160

John McAfee (@officialmcafee), TWITTER (Jan. 3, 2019, 2:24 PM), https://twitter

.com/officialmcafee/status/1080953136985133062?s=21.

161

Crothers, supra note 158.

162

Id.

163

Press Release, U.S. Att’y’s Off., Middle Dist. of Florida, “Queen of Tax Fraud”

Resentenced to 21 Years in Prison (Mar. 6, 2015), https://www.justice.gov/usao-mdfl/pr

/queen-tax-fraud-resentenced-21-years-prison [https://perma.cc/E3CL-6RBA].

164

Wilson’s federal indictment included fifty-seven counts. See U.S. v. Wilson, 593 F.

App’x 942 (11th Cir. 2014), aff’d in part, remanded in part, 649 F. App’x 827 (11th Cir.

2016), cert. denied, 137 S. Ct. 1064 (2017).

165

Susan Taylor Martin, From the ‘Tax Fraud Queen’ to the $980,000 Refund, Tampa

Bay Is a Hotbed of Tax Scammers, T

AMPA BAY TIMES (Feb. 22, 2019), https://www

.tampabay.com/business/from-the-tax-fraud-queen-to-the-980000-refund-tampa-bay-is-a

-hotbed-of-tax-scammers-20190222/ [https://perma.cc/QXM9-RYPD].

166

Robert W. Wood, Queen of Tax Fraud Gets 21 Year Prison Term — for the Second

Time, FORBES (Mar. 6, 2015, 1:26 AM), https://www.forbes.com/sites/robertwood/2015/03

336 OREGON LAW REVIEW [Vol. 99, 301

Neither McAfee nor Wilson is a low-income taxpayer who engaged

in a small amount of cheating or made a small mistake. Wilson was

sentenced to twenty-one years in her plea bargain and was ordered to

forfeit over $2,000,000 in stolen proceeds.

167

McAfee was indicted in

October 2020 for tax evasion and willful failure to file tax returns.

168

It is understandable that some would take a just deserts approach:

the idea that those who (like McAfee and Wilson) post things publicly

are inviting scrutiny and deserve what they get. Indeed, Americans

seem to enjoy stories of stupid criminals and human folly.

169

These

stories proliferate on blogs, in news media, and in books.

170

It stands

to reason, therefore, that people may enjoy stories of taxpayers “dumb”

enough to tip their tax hand to the IRS on social media.

171

In my view,

our government agencies should resist the apparently very human urge

to laugh at human misjudgment. Our revenue agency ought to direct its

resources toward serious and systemic tax evasion and toward reducing

the tax gap.

I do not find McAfee or Wilson to be sympathetic examples either;

while both cases involve social media bragging, they run much deeper

than that. My privacy concerns do not extend to those who choose to

use social media to openly taunt the IRS or encourage the shirking of

/06/queen-of-tax-fraud-teasing-i-cant-be-caught-gets-21-years-prison-for-the-second-time

/#791bcc0a4d15 [https://perma.cc/PDD5-2ZZY].

167

Press Release, U.S. Att’y’s Off., supra note 163.

168

Press Release, U.S. Dep’t of Just., John McAfee Indicted for Tax Evasion (Oct. 5,

2020), https://www.justice.gov/opa/pr/john-mcafee-indicted-tax-evasion [https://perma.cc

/6GWA-AFEB]. McAfee was arrested in Barcelona, Spain, as he was getting ready to board

a flight to Istanbul. Following his arrest, McAfee tweeted from prison: “I am charged with

tax fraud but my only crime is refusing to file returns - a misdemeanor. Fraud is lying on

your tax returns. How could I have lied if I have said nothing?” John McAfee

(@officialmcafee), T

WITTER (Oct. 10, 2020, 12:00 PM), https://twitter.com/officialmcafee

/status/1315004213638897665.

169

See, e.g., DARWIN AWARDS, http://darwinawards.com/rules/ [https://perma.cc/333W

-5QK9]. The Darwin Awards is a “humor” site and a community effort published by Wendy

Northcutt, who has compiled the stories into a book series titled “The Darwin Awards”

(Dutton Books). Id. At the time of this writing, there were four books in the series. The

Darwin Awards website states, “One should not be ashamed of laughing over the misfortune

of others.” Id.

170

See, e.g., TOM NICK COCOTOS, WEIRD BUT TRUE! STUPID CRIMINALS: 150

BRAINLESS BADDIES BUSTED, PLUS WACKY FACTS (2012); Andy Simmons & Priscilla

Torres, The Fifteen Unluckiest Dumb Criminals Ever, R

EADER’S DIG. (Sept. 18, 2019),

http://www.rd.com/funny-stuff/dumb-criminals-unlucky/ [https://perma.cc/8BES-ZVDJ].

171

Similarly, the state of Washington made an example of the individual who publicly

posted an offer to trade marijuana for EBT cards. WASH. ST. DEP’T OF SOC. & HEALTH

SERVS., supra note 63. All three of these examples (McAfee, Wilson, and the EBT trader)

involve criminal acts rather than civil infractions.

2021] #Audited: Social Media and Tax Enforcement 337

tax obligations. The IRS would ignore this behavior at its own peril, as

researchers find a strong link between tax compliance and tax

morale.

172

If taxpayers view tax evasion taunts on social media and

believe that those individuals are “getting away with it,” that

undermines goals of tax compliance.

173

Moreover, any criminal investigations of these two individuals

likely did not emanate from the internet bragging—rather, the bragging

(or taunting) seems to have come after the individuals were under

scrutiny by the authorities. My concerns are with much more subtle

social media users. For example, ordinary photos scanned by an

algorithm might prompt an audit; the algorithm might predict that,

based on these photos, an individual earns more money than he or she

reported, or that a child does not live in the taxpayer’s house for the

period required in connection with a particular credit that he or she

claimed.

In addition to my concern that the use of social media mining might

disproportionately harm low-income taxpayers, I wonder if the

agency’s choice to engage in social media mining might impact certain

personality types of taxpayers more than others. For example, it seems

that social media mining would likely locate individuals who engage

in bluster, targeting for scrutiny those who are loud, flashy, or

indiscreet, even though those individuals’ tax compliance may be no

better or worse than quiet individuals. Another concern is that it might

punish those who are cultivating a false image of themselves online.

174

In a survey of 2000 British social media users, more than 75% of people

admitted to lying about themselves on Twitter and Facebook.

175

There

is a rich body of social science literature examining how and why

172

See Kornhauser, supra note 144, at 139 (“Higher tax morale correlates with

higher tax compliance.”). Kornhauser’s article provides an extensive bibliography of tax

compliance literature.

173

Kornhauser notes that “[n]on-compliance among other taxpayers can decrease an

individual’s own tax morale and compliance.” Id. at 139–40 (citing Bruno S. Frey & Benno

Torgler, Tax Morale and Conditional Cooperation, 35 J. COMP. ECON. 136 (2007)).

174

Houser and Sanders address the problem of potentially false data in their article. See

Houser & Sanders, supra note 14, at 841–42 (citing Minas Michikyan, Jessica Dennis &

Kaveri Subrahmanyam, Can You Guess Who I Am? Real, Ideal, and False Self-Presentation

on Facebook Among Emerging Adults¸ 3 E

MERGING ADULTHOOD 55, 60 (2015)).

175

Lisa Vaas, Over 75% of People Lie on Social Media, NAKED SECURITY: SOPHOS

(Apr. 7, 2016), https://nakedsecurity.sophos.com/2016/04/07/over-75-of-people-lie-on-social

-media/ [https://perma.cc/QB3T-T8EM] (finding men more likely to lie on social media than

women, with 43% of men polled admitting to fabricating facts about themselves); see also

Cortney S. Warren, How Honest Are People on Social Media?, P

SYCH. TODAY (July

30, 2018), https://www.psychologytoday.com/us/blog/naked-truth/201807/how-honest-are-

people-social-media [https://perma.cc/WL8W-KFBC] (describing the same British survey).

338 OREGON LAW REVIEW [Vol. 99, 301

people lie online, as well as perceptions of how often other people

lie.

176

This tendency for people to lie or exaggerate surely spans all ages,

but there appears to be a generational element to it as well. Social

scientists have found that young adults “with a less coherent sense of

the self and lower self-esteem reported presenting their false self on

Facebook to a greater extent.”

177

Whether young or old, people may

have a wide variety of incentives and motivations to cultivate a false

online image. They may post pictures of someone else’s expensive car.

In some cases, a photo may not tell the full story. For example, a low-

income individual may post photos of a vacation at an expensive resort,

which might suggest to the IRS that the individual has significant

disposable income—but perhaps this individual’s parents paid for the

entire trip. Do we want an algorithm to point the IRS in these directions,

forcing individual taxpayers to prove that someone else took them

someplace expensive and paid their way?

In this regard, it is likely data mining will pick up false hits and

minor indiscretions that will direct the IRS down a path that is

either negligible (in the case of a low-income taxpayer who has

committed a relatively minor tax indiscretion, either intentionally or

unintentionally) or useless (in the case of those presenting a false image

or those posts that the algorithm misinterprets).

178

Other government

agencies that engage in social media investigations have cited this as a

downside of using automated tools.

179

Thus, it is foreseeable that the

IRS’s use of automated mining may necessitate the resource-intensive

manual review of many accounts that turn out to be false hits or not

worth pursuing; in that regard, it is possible that automation may not

be as cost-saving as one would assume. Kimberly Houser and Debra

Sanders raised similar concerns about the accuracy of IRS data mining,

noting, “big data results are based on correlation, not causation, and it

is inappropriate to judge people based on correlation; just because

176

See, e.g., Michelle Drouin et al., Why Do People Lie Online? “Because Everyone

Lies on the Internet,” 64 COMPUTS. HUM. BEHAV. 134 (2016) (examining online deception

across four different online venues, including social media).

177

Michikyan et al., supra note 174, at 55.

178

The IRS RFI mentions this possibility, though does not elaborate on how it would

prevent it: “the IRS will also be mindful that frequently information posted on social media

and the internet may be wrong or misleading.” IRS RFI, supra note 2.

179

See SUPPLEMENTAL NUTRITION ASSISTANCE PROGRAM, supra note 58, at 23

(“[O]fficials in one state noted that the automated tools have placed an excessive demand

on staff because they had to sift through the many false-positive leads that were generated.”).

2021] #Audited: Social Media and Tax Enforcement 339

people share characteristics or interests does not mean that they will

have similar tax compliance behavior.”

180

Returning to the example of Predictim, the now-defunct online

service that used artificial intelligence screening to create a predictive

profile of potential babysitters, computer-generated scans sometimes

produce ambiguous results. “When one babysitter’s scan was flagged

for possible bullying behavior, the unnerved mother who requested it

said she couldn’t tell whether the software had spotted an old movie

quote, song lyric or other phrase as opposed to actual bullying

language.”

181

Similarly, Facebook has worked to fine-tune its social

suicide prediction screening: Mason Marks cites the example of how in

its early efforts Facebook picked up false positives from likely benign

social media posts such as “Ugh, I have so much homework I just

wanna kill myself.”

182

At the same time, the IRS’s use of social media data mining

presumably would spare those who are quiet—those who are

sophisticated enough to maintain online privacy, who know what not

to post, those who are so down-and-out that they lack access to the

internet, or those who lack the instinct to post about their lives on social

media. A sophisticated individual will follow the advice of privacy law

scholar Joshua Fairfield, who advises that one way for a device user to

protect one’s privacy is to “provide false information where possible,

to make the algorithms less sure.”

183

I can imagine instances in which an algorithm might also target those

who maintain a social media presence but are silent as to certain aspects

of their lives. Imagine a taxpayer who has qualifying children she

claims for the EITC and child tax credit. What if she maintains an active

Facebook page but never posts pictures of her children? This is not a

farfetched scenario: many people choose not to post photos of their

children due to privacy concerns or simply because their children ask

them not to do so. Based on the absence of photos, would an algorithm

suggest that the IRS select this taxpayer’s return for examination and

require substantiation for the qualifying child claims, even if the

existing databases (e.g., the Dependent Database) had not prompted it?

180

Houser & Sanders, supra note 14, at 871.

181

Harwell, supra note 71.

182

Marks, supra note 25, at 109.

183

Joshua Fairfield, 7 Things I Teach My Kids About Privacy, MEDIUM (Sept. 9, 2020),

https://medium.com/@fairfieldj/7-things-i-teach-my-kids-about-privacy-24372552cde0

[https://perma.cc/5DTP-42T4] (explaining that, as device users, we lack the ability to protect

our data but can take certain steps to protect our privacy).

340 OREGON LAW REVIEW [Vol. 99, 301

Another concern is the proliferation of fake social media accounts,

including for nefarious purposes.

184

There have been publicized cases

of criminals impersonating individuals on Facebook in order to scam

unsuspecting individuals.

185

Would this affect the social media mining

algorithm, leading to time-consuming false positives?

These are but a few suggestions of how social media algorithms

might create red herrings for the IRS in many directions, directing

scarce resources to dead ends rather than prioritizing human analytical

work toward larger dollar cases. Are there other, more effective ways

to use social media as a tax compliance tool? Returning to the extreme

examples of internet braggadocio, the IRS Whistleblower Program

provides an example that might be instructive to draw upon in this

regard.

B. Whistleblowers: Should the IRS Encourage Social Media

Snitching, and Under What Circumstances?

Recall the story of the woman in Pennsylvania who brazenly posted

an offer on Facebook to trade her SNAP card for cash.

186

Social media

was her downfall, but she wasn’t caught by an algorithm or even by an

agency employee conducting a manual search.

187

Something old-

fashioned happened: someone snitched on her.

188

A member of the

public saw the post, recognized the illegal nature of it, and contacted

the agency to report it.

189

This example underscores something about

human nature: if people are morally outraged by something, they will

speak up. Thus, it is possible that social media will serve as a tool for

184

See Arun Vishwanath, Habitual Facebook Use and Its Impact on Getting Deceived

on Social Media, 20 J.

COMPUTER-MEDIATED COMM. 83 (2015) (describing how habitual

Facebook use may increase the user’s vulnerability to phishing attacks on social media

sites).

185

See, e.g., Jack Nicas, Facebook Connected Her to a Tattooed Soldier in Iraq. Or

So She Thought, N.Y. T

IMES (July 28, 2019), https://www.nytimes.com/2019/07/28

/technology/facebook-military-scam.html [https://perma.cc/UR8E-PVWR] (“[There are]

two sides of a fraud that has flourished on Facebook and Instagram, where scammers

impersonate real American service members to cheat vulnerable and lonely women out of

their money.”).

186

Following the tip from a member of the public, investigators arrested Tanya Keenan

Mac after she traded her EBT card for heroin. Kristina Papa, Food Stamp Facebook Post

Leads to Another Arrest, WNEP (Apr. 4, 2017, 5:50 PM), https://www.wnep.com/article

/news/local/lycoming-county/food-stamp-facebook-post-leads-to-another-arrest/523-8c2dc

cca-b352-42bb-9e37-4eafa4fdf6d9 [https://perma.cc/Y49Y-972U].

187

Id.

188

Id.

189

Id.

2021] #Audited: Social Media and Tax Enforcement 341

enforcement whether the IRS chooses to mine it or not. Not every

instance of internet braggadocio is as outrageous as the examples of

John McAfee and Rashia Wilson or even as overt as that of the woman

in Pennsylvania. Most social media leads will be subtler and may be

useful only in conjunction with other information known to the viewer.

In his 1996 article suggesting ways to improve public perceptions of

the tax system, Joshua Rosenberg proposed ways to encourage and

even financially incentivize the public to report tax avoidance to the

authorities, arguing that fostering communication (rather than

polarization) with the IRS may improve attitudes toward tax

compliance.

190

Along those lines, perhaps it might be more appropriate

to leave social media monitoring to the public and not have the IRS

initiate it. In that framework, the IRS is not doing the frontline social

media mining, which eliminates the unwelcome “Big Brother” feeling

of the government monitoring your posts.

The IRS already has some mechanisms in place for members of the

public to report tax noncompliance. Anyone can report a “suspected tax

law violation” on IRS Form 3949-A, which can be submitted

anonymously.

191

The form provides a place to report the name of the

alleged wrongdoer, and the informant can tick one or more of twenty

boxes to describe the alleged violations. These boxes range from very

common types of civil noncompliance (failure to file a return, failure

to pay tax, EITC, unreported income) to more serious criminal

allegations (narcotics income, kickback, organized crime, corruption).

In addition to this information referral form, which anyone can use

to report a variety of suspected misconduct, the IRS has a statutory

whistleblower program. Internal Revenue Code § 7623, as originally

enacted, was an informant claim program authorizing the Secretary of

Treasury to pay financial awards as necessary for “(1) detecting

underpayments of tax,” or “(2) detecting and bringing to trial and

punishment persons guilty of violating the internal revenue laws or

190

Joshua D. Rosenberg, The Psychology of Taxes: Why They Drive Us Crazy, and How

We Can Make Them Sane, 16 VA. TAX REV. 155 (1996).

191

IRS, FORM 3949-A, INFORMATION REFERRAL (2016). IRS procedures for screening

these forms are described in the Internal Revenue Manual, though parts are redacted. IRM

3.28.2 (Aug. 1, 2020).

342 OREGON LAW REVIEW [Vol. 99, 301

conniving at the same.”

192

Section 7623(b), added in 2006,

193

created

a formalized whistleblower program, providing that a whistleblower

shall “receive as an award at least 15 percent but not more than 30

percent of the proceeds collected as a result of the action . . . or from

any settlement in response to such action,” with the amount of the

award determined based upon “the extent to which the individual

substantially contributed to such action.”

194

The 2006 amendment also

mandated the creation of an IRS Whistleblower Office to implement

the program.

195

Under the IRS Whistleblower Program, an award applies only in

cases in which (1) an individual taxpayer’s gross income exceeds

$200,000 in the year at issue; or (2) the proceeds in dispute exceed

$2,000,000.

196

Thus, Congress designed the program to incentivize tips

or reports of significant evasion or of evasion by taxpayers with

significant income.

Notably, the award is decreased in cases “based principally on

information disclosed in certain public sources . . . .”

197

This limitation

192

I.R.C. § 7623(a). Internal revenue informant laws have been in effect since 1867, but

the significant revisions that created the whistleblower program were enacted in 2006.

See IRS, History of the Whistleblower/Informant Program (June 18, 2020), https://www.irs

.gov/compliance/history-of-the-whistleblower-informant-program [https://perma.cc/8HF3

-HUY2].

193

Tax Relief and Health Care Act of 2006, Pub. L. No. 109-432, div. A, tit. IV, § 406,

120 Stat. 2958.

194

I.R.C. § 7623(b).

195

Tax Relief and Health Care Act of 2006 § 406(b).

196

I.R.C. § 7623(b)(5). The most prominent IRS whistleblower is Bradley Birkenfeld,

the UBS banker who divulged information to the IRS about Swiss banking schemes that

amounted to criminal tax evasion. David Kocieniewski, Whistle-Blower Awarded $104

Million by I.R.S., N.Y. T

IMES (Sept. 11, 2012), https://www.nytimes.com/2012/09/12/

business/whistle-blower-awarded-104-million-by-irs.html [https://perma.cc/Q37C-6GAB].

Birkenfeld, himself a subject of the criminal investigation and who served two and a half

years in prison for his role in the scheme, made use of the IRS Whistleblower Program and

received the largest ever award, $104 million. Id. The information he provided to the IRS

yielded a lot of fruit: thousands of names of U.S. offshore bank account owners were

provided to the IRS, and thousands of other offshore bank account holders voluntarily

disclosed their accounts as part of an amnesty program that resulted.

Id. The IRS estimated

that Birkenfeld’s information “helped recover more than $5 billion in unpaid taxes.”As the

article notes, whistleblower awards constitute gross income, so the IRS ultimately receives

part of the award back in tax revenue. Id.

197

IRS, PUBLICATION 5241, WHISTLEBLOWER PROGRAM FISCAL YEAR 2018 ANNUAL

REPORT TO CONGRESS 4 (2019), https://www.irs.gov/pub/irs-prior/p5241--2019.pdf [https:

//perma.cc/3HSB-N2AE]; see also I.R.C. § 7623(b)(2). The award is also decreased when

the whistleblower “planned and initiated the actions” that led to the tax law violations. I.R.C.

§ 7623(b)(3).

2021] #Audited: Social Media and Tax Enforcement 343

is likely to become relevant if a whistleblower’s basis for the claim is

information he or she saw on a taxpayer’s Facebook or Twitter page.

While most tips from social media would be unlikely to meet the

criteria of the formalized whistleblower program, the IRS has fixed a

general framework for individuals to report noncompliant taxpayers.

Instead of pursuing a third-party tool for social media mining, the IRS

could simply leave the online snooping to the public, and the agency

can follow its existing procedures in pursuing any social media-related

tips that are reported on Form 3949-A or whistleblower claims.

C. Proposals for Setting IRS Policies on Social Media Mining:

Balancing Modern Enforcement Techniques with a Taxpayer

Right to Privacy

1. The IRS Should Clarify Its Understanding of the Taxpayer “Right

to Privacy”

As a starting point, the IRS ought to define more clearly its

interpretation of a taxpayer’s right to privacy. Michael Hatfield’s

scholarship provides examples illustrating the breadth of personal

information that may be tax relevant.

198

Hatfield warned that

technological advances, including the use of big data mining, would

undo the privacy protection previously afforded to individuals when

the IRS faced a structural inability to review all the information

available to it.

199

Notions of privacy are rapidly changing. Perhaps it is not even

possible for the IRS to define its conceptualization of taxpayer privacy.

That would be a legitimate position—if that is the case, it ought to say

so explicitly. In part parroting the statutory language of § 7602(e), the

IRS website explanation of the taxpayer right to privacy states the

agency “should not seek intrusive and extraneous information about

your lifestyle during an audit if there is no reasonable sign that you

198

Hatfield, supra note 14, at 631 (referencing the potential tax relevance of sleeping

arrangements, marital discord, therapy sessions, and health and social club memberships,

among other things); see also Michael Hatfield, Taxation and Surveillance: An Agenda,

17 Y

ALE J.L. & TECH. 319, 321 (2015) (describing instances of how the IRS is entitled to

collect information about individuals’ hobbies, reading preferences, religious affiliation,

travel plans, weight and doctor’s recommendations about it, abortion, sterilization, or gender

identity disorder).

199

Hatfield, supra note 14, at 631–32 (“[G]iven how much personal information will be

covered by the coming technology and how much personal information is potentially tax

relevant, it is hard to have anything but a dystopian vision of this future . . . .”).

344 OREGON LAW REVIEW [Vol. 99, 301

have unreported income.”

200

However, most of what is posted on social

media is extraneous to one’s income situation. If the IRS is routinely

surveilling social media at large (even if by algorithm as opposed to

manually), rather than making targeted searches after finding signs of

unreported income, then it is seemingly violating its own principle.

2. Increase Transparency of Audit Techniques

Relatedly, if the IRS does not believe that taxpayers have a right to

privacy in their social media posts, and if it intends to use automated

techniques to trawl all forms of social media for possible enforcement

leads, it ought to state that intention clearly and explicitly. At least then

the public will be on explicit notice that the IRS is watching their social

media. Just as posted signs warn motorists of traffic cameras on roads

or notify the public of security cameras at tourist sites and stores,

taxpayers should be on notice that the IRS is monitoring their social

media activity. Further, the IRS ought also to provide specifics as to its

methods: whether the mining is manual or automated; whether the

agency is searching indiscriminately or only with cause; whether it is

doing so in civil cases or only for criminal investigations; and whether

it is doing so in connection with examinations, collections, or both.

This information ought to be prominently displayed on the IRS

website. The IRS could undertake a public relations campaign to this

effect, using social media as its own tool for spreading the word to

taxpayers. If it wishes, it might even openly encourage people to report

one another when they see incriminating information on other people’s

social media. Imagine such a tweet from the IRS: “Do your patriotic

duty: retweet suspected tax noncompliance! #TaxGap @IRStaxpros

#taxcompliance.” If that is how the IRS wishes to operate, it should

make that known to the public so that people can proceed accordingly.

3. Limit Social Media Investigations to Manual Searches Rather t

han

Automated, and Define Limits in the Internal Revenue Manual

If the IRS pursues social media mining as part of its enforcement

strategy, it will face decision points in setting the contours. A critical

first question is to what extent the mining should be done by humans

as opposed to by algorithms. If the mining is done by algorithms, a

200

Taxpayer Bill of Rights 7: The Right to Privacy, IRS (Jan. 22, 2021), https://www

.irs.gov/newsroom/taxpayer-bill-of-rights-7-the-right-to-privacy-0 [https://perma.cc/2CH5

-KR7D].

2021] #Audited: Social Media and Tax Enforcement 345

second question is whether mining should be limited only to current

investigations or used more widely to cast the broadest possible net.

Software programs for data mining produce false positives and are

potentially discriminatory. Of course, the same risks are inherent in

humans performing manual searches on social media. But humans can

play a role in procedural protections, especially if trained in how to

conduct searches in a way that minimizes false positives. In the context

of whether the Social Security Administration should explore greater

use of social media in its investigatory work, Social Security

Commissioner Michael J. Astrue told members of Congress that, in his

view, professionally trained fraud investigators should be the ones

evaluating circumstantial evidence of fraud, not administrative law

judges or other employees: “Social media sites are not exactly clear and

reliable evidence . . . Facebook puts up phony websites under my name

all the time.”

201

Additionally, humans face constraints that computers do not: for

one, they have only a limited number of work hours in the week.

Humans do not have time to engage in widespread social media mining

on a random basis, nor would it be cost efficient for them to do so.

Thus, it would be logical to limit employee use of manual searches to

the context of existing enforcement cases. This would be most

consistent with the limits imposed by § 7602(e), providing that the IRS

should not engage in broad fishing expeditions based solely on signs of

financial status but can conduct lifestyle audits once the agency finds

“a reasonable indication that there is a likelihood of such unreported

income.”

202

Authorizing specially trained employees to undertake manual

searches of social media in preexisting enforcement cases would be

the least harmful alternative. There is a humanity to manuality that is

not inherent in automated searches: it opens the door to mercy and

201

Robert Pear, On Disability and on Facebook? Uncle Sam Wants to Watch What

You Post, N.Y. TIMES (Mar. 10, 2019), https://www.nytimes.com/2019/03/10/us/politics

/social-security-disability-trump-facebook.html [https://perma.cc/BM7V-FA9D]. Pear quotes

Astrue’s remarks in the 2019 article; the remarks were made in 2012 while Astrue was

serving as Social Security Commissioner. In his testimony at the hearing, Astrue articulated

the possibility that angry ex-spouses may post false information on Facebook. The Social

Security Administration: Is It Meeting Its Responsibilities to Save Taxpayer Dollars

and Serve the Public?: S. Hearing 112-751 Hearing Before the Comm. on Fin., 112th

Cong. (2012) (statement of Hon. Michael J. Astrue, Commissioner, Social Security

Administration). He also expressed a concern that “[i]f you allow broad social media access

on government time, I think that becomes an enormous suck on productivity.” Id.

202

I.R.C. § 7602(e).

346 OREGON LAW REVIEW [Vol. 99, 301

discretion in the taxpayer’s favor. There is, however, still the risk of

employee bias with manual searches. If the IRS were to adopt this

technique, it ought to develop policies on when social media

investigations are appropriate and what types of evidence might be

probative. The IRS can incorporate these policies into the Internal

Revenue Manual. Most importantly, these tools ought not to be used to

unduly scrutinize poor people.

For example, with respect to collection enforcement, it may be

appropriate for an Offer Specialist to manually search social media

while verifying the financial statements that were made by the

applicant under penalties of perjury. Perhaps an Offer Specialist’s

findings should weigh adversely only if a material misrepresentation is

found and documented in writing in the file, and after the applicant has

had a chance to respond to specific questions about the information

viewed online. Imagine a manual social media search that appears to

uncover an asset of significant value not reported on the Offer in

Compromise financial statement: the Offer Specialist should be

allowed to make further inquiries and document the findings and the

applicant’s response as part of the administrative record. For example,

if the Specialist sees a photograph of the taxpayer wearing a very

expensive wristwatch that was not reported on the asset section of the

financial statement, it is reasonable that the Specialist be permitted to

ask the taxpayer about this. On the other hand, the Specialist should not

be permitted to make value-laden inquiries as to the applicant’s

expenses. For example, taxpayers submitting an Offer in Compromise

are permitted to use a standard expense dollar amount to account for

food, clothing, and certain other household items based on the size of

the household; the IRS policy is to permit the standard figure without

questioning the amount actually spent, and the taxpayer is not required

to substantiate the allowable standard figure.

203

Therefore, if the

taxpayer has claimed the standard figure, the Offer Specialist should be

barred from asking the taxpayer to increase the offer’s dollar amount

on the basis that the taxpayer has been spotted on social media smoking

cigarettes, eating steak, or wearing designer-brand clothes.

203

See Collection Financial Standards, IRS (Nov. 19. 2020), https://www.irs.gov

/businesses/small-businesses-self-employed/collection-financial-standards [https://perma

.cc/3TGE-JS9B].

2021] #Audited: Social Media and Tax Enforcement 347

4. If Automated Social Media Mining Is Used, Implement Use of Pre-

Examination Soft Letters to Nudge Taxpayers Detected by the

Algorith

If the IRS proceeds with using social media mining, it should do so

as part of an announced compliance campaign. Rather than moving

straight to opening examinations based on results, as a first step, the

IRS should send selected taxpayers a “soft letter” as a nudge to improve

compliance. The idea behind such an initiative is to let taxpayers know

that the IRS is undertaking a new enforcement technique or enhancing

enforcement of a particular type. The IRS refers to these as

“educational letters,” and there is precedent for broad use of these

letters in targeted areas of suspected noncompliance. In recent years,

the IRS has sent soft letters to cryptocurrency holders,

204

offshore

account holders,

205

and taxpayers known to have engaged in specific

types of transactions that the IRS suspects might be abusive.

206

The Taxpayer Advocate Service conducted a study in which it sent

educational letters to taxpayers whose returns were not selected for

examination, but nonetheless appeared to have erroneously claimed the

EITC on their 2014 tax return.

207

The study found that the letters were

effective in improving compliance the following year in some

particular contexts, with taxpayers not making the same mistake the

next year.

208

During her time as National Taxpayer Advocate, Nina

204

See, e.g., I.R.S. News Release IR-2019-132 (July 26, 2019). Taxpayer Advocate Erin

Collins recently criticized the scope of these letters, calling it “disturbing” that these letters

imposed “unreasonable burdens on [taxpayers] outside the protection of an examination.”

AT’L TAXPAYER ADVOC., OBJECTIVES REPORT TO CONGRESS FISCAL YEAR 2021, at 80

(2020).

205

See, e.g., Kelly Phillips Erb, IRS to Target S Corporations, Foreign Disclosures in

New Campaigns, FORBES (July 24, 2019, 2:15 PM), https://www.forbes.com/sites/kelly

phillipserb/2019/07/24/irs-to-target-s-corporations-foreign-disclosures-in-new-campaigns

/#78f1f4313a38 [https://perma.cc/CE98-9U2H].

206

See, e.g., Jay Adkisson, IRS Sends Second Wave of Soft Letter Warnings to Certain

Captive Owners for Abusive Microcaptive Transactions, FORBES (Aug. 8, 2020, 9:48 PM),

https://www.forbes.com/sites/jayadkisson/2020/08/08/irs-sends-second-wave-of-soft-letter

-warnings-to-certain-captive-owners-for-abusive-microcaptive-transactions/#7b14cd8713ee

[https://perma.cc/6746-857W].

207

See NAT’L TAXPAYER ADVOC., STUDY OF SUBSEQUENT FILING BEHAVIOR OF

TAXPAYERS WHO CLAIMED EARNED INCOME TAX CREDITS (EITC) APPARENTLY IN

ERROR AND WERE SENT AN EDUCATIONAL LETTER FROM THE NATIONAL TAXPAYER

ADVOCATE, 2 ANNUAL REPORT TO CONGRESS 32, 33 (2016).

208

Id. The study found the educational letters were particularly effective with respect to

educating taxpayers on the relationship requirement of the EITC but less effective with

respect to taxpayers who claimed a child that was also claimed by another taxpayer.

Extrapolating from the study, some types of noncompliance may be more easily corrected

by an educational letter than others. See also NTA Blog, EITC – How a Simple Educational

348 OREGON LAW REVIEW [Vol. 99, 301

Olson emphasized that, in order to be salient and effective, IRS letters

or notices should be specific in content and tailored to the most relevant

issues to the taxpayers.

209

To be most effective, an educational letter in

the social media context should address the specific mistake that is

suspected, rather than be framed as a general reminder to comply with

taxes.

While I do not wish to see the IRS engage in widescale automated

social media mining, sending educational letters to taxpayers based

upon suspicious findings is a more reasonable first step than

immediately opening an examination. The letter should specifically

reference social media. Sending such a letter serves the goal of

transparency, as it puts the taxpayer on notice that the IRS is surveilling

their online activity and, ideally, will nudge the taxpayer to correct his

or her behavior.

210

5. Sharply Define the Social Media Mining Criteria, Using It t

Target Only the Most Egregious Noncompliance

I find social media mining in tax enforcement cringeworthy at any

income level, and my first preference is to keep the IRS out of this

realm. However, if the IRS proceeds with this initiative, my alternative

preference is for the agency to set its sights either on the most egregious

types of tax noncompliance (regardless of income level) or on the

highest earners, who have been the subject of a declining audit rate in

the last decade.

211

Letter Can Help Avert Noncompliance, TAXPAYER ADVOC. SERV. (Feb. 28, 2018), https:

//www.taxpayeradvocate.irs.gov/news/ntablog-eitc-how-a-simple-educational-letter-can

-help-avert-noncompliance/ [https://perma.cc/BGA9-KAVM].

209

NTA Blog, The IRS Should Redesign Its Notices Using Psychological, Cognitive,

and Behavioral Science Insights to Protect Taxpayer Rights, Enhance Taxpayer

Understanding, and Reduce Taxpayer Burden, T

AXPAYER ADVOC. SERV. (April 3, 2019),

https://www.taxpayeradvocate.irs.gov/news/ntablog-the-irs-should-redesign-its-notices

-using-psychological-cognitive-and-behavioral-science-insights-to-protect-taxpayer-rights

-enhance-taxpayer-understanding-and-reduce-taxpayer-burden/ [https://perma.cc/W97B

-UFFF] (referring to notices in general, not educational letters in particular).

210

As governments in many other countries have done, the IRS has borrowed upon

behavioral science, psychology, and cognitive science in connection with its tax compliance

research; the IRS can use insights gleaned from its studies to design a social media soft

letter and determine which taxpayers should receive it. See T

AXPAYER ADVOC. SERV.,

LITERATURE REVIEW: IMPROVING NOTICES USING PSYCHOLOGICAL, COGNITIVE, AND

BEHAVIORAL SCIENCE INSIGHTS, 2 ANNUAL REPORT TO CONGRESS 194 (2018).

211

See, e.g., Keith Fogg, IRS Large Case Examination Rate Collapses, PROCEDURALLY

TAXING (July 13, 2020), https://procedurallytaxing.com/irs-large-case-examination-rate

-collapses/ [https://perma.cc/VTL2-ECTD] (citing statistics from IRS 2019 Databook).

2021] #Audited: Social Media and Tax Enforcement 349

For example, could the IRS use social media mining to screen for

potential tax protestors? It could set its data mining algorithm to search

for social media posts that repeat or promote frivolous tax arguments,

such as those compiled by the IRS on its website.

212

With respect to pursuing wealthy tax cheats, the IRS might wish to

look to the examples of state revenue agencies and foreign revenue

agencies using high-tech methods in enforcement. The New York State

Department of Taxation is well known for using invasive methods

(including social media) to track individuals who have ties to New

York but claim to reside in lower-tax states on their tax filings.

213

These residency audits are aimed, strategically, at the wealthy as a

means of maximizing the state’s revenue base.

214

Greece made

headlines when it used police helicopters and satellite images from

Google Earth to locate home swimming pools as part of a large-scale

crackdown on tax evasion and unreported income.

215

From a cost-benefit perspective, it makes sense to aim these

initiatives at the wealthy if the revenue recouped from the positive

results will offset the costs of chasing false leads. To some, these

initiatives may create “feel-good” stories with the hope that such stories

will boost tax morale (and thus boost individual compliance) among

the larger public.

216

Personally, I do not like the idea of the IRS

tracking any class of individuals’ movements electronically—

normatively, this is more privacy than I wish for the public to sacrifice

in the name of revenue collection

217

—so I do not favor this approach

212

The Truth About Frivolous Tax Arguments, IRS (May 1, 2020), https://www.irs.gov

/privacy-disclosure/the-truth-about-frivolous-tax-arguments-introduction [https://perma.cc

/958W-MGGX].

213

See, e.g., Robert Frank, Tax Collectors Chase Rich New Yorkers Moving to Low-Tax

States. Auditors Inspect Cell Records, Even Your Dog’s Vet Bills, CNBC (Mar. 8, 2019,

7:05 AM), https://www.cnbc.com/2019/03/08/tax-collectors-chase-rich-new-yorkers-moving

-to-low-tax-states.html [https://perma.cc/2EHE-ECL4].

214

Id. (“New York can’t afford to lose many millionaires or billionaires. The top 1

percent of earners pays 46 percent of the state’s income taxes . . . .”).

215

Daniel Steinvorth, Finding Swimming Pools with Google Earth: Greek Government

Hauls in Billions in Back Taxes, SPIEGEL INT’L (Aug. 2, 2010, 2:48 PM), https://www

.spiegel.de/international/europe/finding-swimming-pools-with-google-earth-greek-govern

ment-hauls-in-billions-in-back-taxes-a-709703.html [https://perma.cc/NCA5-V553]

(describing how these tactics revealed that the suburbs of Athens had 16,974 swimming

pools, rather than the 324 that had been reported).

216

Studies suggest that taxpayer compliance is directly correlated to perceived

compliance of others. See, e.g., Alm et al., supra note 104, at 297.

217

Recall that the IRS states that taxpayers have the right to expect enforcement action

will be “no more intrusive than necessary.” In my opinion, using data mining and

technological surveillance to track the personal lives of taxpayers without a specific

350 OREGON LAW REVIEW [Vol. 99, 301

even if targeted to the wealthy. But to the extent the IRS wishes to adopt

these methods, I prefer targeting tax protestors, or the wealthy, to the

idea of training an algorithm to tease out whether a poor person earns

a little bit of extra cash on the side or whether a person in a $25 per

month installment agreement could afford to pay a slightly higher

amount but for the fact that she is smoking a pack of cigarettes a day.

Targeting specific groups other than the poor would also serve as a

counterweight to the fact that the lowest-income taxpayers are already

subject to such a high audit rate. If nothing else, the agency’s

enforcement efforts would be spread among a larger group of taxpayers

(instead of increasing the focus on the poorest).

6. Use Social Media Mining Only at Taxpayer’s Request, as a Met

hod

of Dispute Resolution

As mentioned in the introduction, the RFI also raises the prospect of

using a social media tool to protect taxpayers. While I do not favor the

IRS searching individual taxpayers’ social media in any capacity, I

would be curious to see how the IRS might use it to protect taxpayers:

trading off taxpayer privacy only for good, rather than for

enforcement.

218

Recall the example of the man in Detroit who was

wrongfully arrested and could have invoked social media in his alibi—

are there similar analogies to be envisioned for taxpayers?

219

Imagine if taxpayers could turn social media in their favor, electing

at their option to use it as a defense or to substantiate their claims in

disputes. For example, IRS underreporter notices are sometimes the

first clue to taxpayers that they are victims of identity theft; because the

notice lists all tax information reported by third parties, the taxpayer

will see if income is wrongly reported by an unknown person who is

working under the taxpayer’s social security number. In those cases,

the taxpayer must contact the IRS and establish that the income listed

on the notice does not belong to the taxpayer. Perhaps in some cases a

taxpayer could use social media posts to demonstrate to the IRS that

they had only one place of employment or that they had no connection

suspicion of tax noncompliance is far more intrusive than tracking wages and financial

transactions, the latter of which are justifiable for a revenue agency.

218

Relatedly, though not specific to social media, see W. Edward Afield, Moving Tax

Disputes Online Without Leaving Taxpayer Rights Behind, 74 TAX LAW. 1 (2020)

(imagining how the IRS can deploy technology to resolve tax disputes in ways that are pro-

taxpayer rights).

219

See supra note 47 and accompanying text.

2021] #Audited: Social Media and Tax Enforcement 351

to the geographic location where the income was reported under their

number.

Perhaps social media could be used (again, I would propose it be

allowed only at the taxpayer’s election) to substantiate an EITC claim

upon examination. The IRS could institute a policy that the agency may

consider a taxpayer’s social media evidence within its discretion; to

protect taxpayers, it could further adopt a policy that no negative

inference can be drawn from the social media sources.

Some individuals have turned to technology (and opted to sacrifice

a bit of their own privacy) as a method of tax planning: as a response

to state residency tax audits, tax-specific compliance tools have

become commercially available. Monaeo sells a so-called personal and

audit defense system, which consists of an app and web interface that

enable users to track and log their days in and out of a state, ensuring

that they do not trip into residency status.

220

In discussing compliance burdens and how cultural norms around

technology and privacy have changed and will continue to evolve,

Michael Hatfield imagines a future in which taxpayers could choose

between two systems depending on their privacy preference.

221

One

option Hatfield sets forth would be to voluntarily sacrifice privacy (by

consenting to data surveillance) and in exchange receive certain tax

benefits; the other option would be to opt out of surveillance and as a

result give up the tax benefit.

222

At least such a regime would put the

taxpayer in control.

III

ROADER IMPLICATIONS FOR REPRESENTING LOW-INCOME

TAXPAYERS IN THE #TMI ERA

In light of the IRS social media request for information, as well as

broader trends signaling the loss of individual privacy, how should tax

practitioners advise their clients? As the director of a low-income

taxpayer clinic, I am most concerned with this vulnerable population

of clients who receive our legal services on a pro bono basis.

220

Frank, supra note 213. The article states that Monaeo said use of its “Personal

Edition” app is up fifty-one percent in 2018 over 2017. See also M

ONAEO, https://monaeo

.com/personal [https://perma.cc/RL64-9USL].

221

Hatfield, supra note 198.

222

Id. at 352. Hatfield emphasizes the need for research as to taxpayer preferences, and

he recognizes that “the tax law of 2040 should be fundamentally different than that of 2015

if revolutionary technologies are to be integrated into its administration.” Id. at 366.

352 OREGON LAW REVIEW [Vol. 99, 301

Increasingly, I see social media usage addressed in a variety of ways

as a continuing legal education (CLE) topic. One such seminar advised

lawyers how to use the internet as an investigative tool; the promotional

flyer included agenda items such as “[f]ind out ‘secret’ ways to ferret

out information from social media profiles” and suggested social

media site navigation as a method to obtain background information

about adverse parties, lawyers, judges, and current and potential

clients.

223

The CLE flyer suggests other uses for lawyers to navigate

social media sites, including to (1) “[F]ind information to attack a party

or witness’s credibility”; (2) “uncover fraud”; and (3) “seek out the

smoking gun.”

224

Other CLE programs include discussion of how the profession’s

ethical rules apply to social media. For example, may an attorney

representing the defendant in a products liability case check the

plaintiff’s social media sites without the plaintiff’s lawyer’s

consent?

225

C. Simon Davidson of law firm McGuire Woods posed

this hypothetical and concluded yes, so long as there is no

“communication” with the party or witness.

226

Davidson cautions,

however, that some state professional regulations “would prohibit

arguably deceptive conduct designed to gain access” to those social

media sites; presumably this would include setting up a fictional

account to “friend” the party on Facebook or “follow” the party on

Twitter.

227

If lawyers are regularly using social media as a tool for opposition

research, what advice are they giving their own clients about how to

manage social media without pitfalls? The obvious advice would be for

clients to simply go dark: no Facebook, no Twitter, no Instagram.

However, that might not be viable advice in the twenty-first century, or

clients may not take it seriously. But it does not hurt to remind clients

that the government may be mining their social media. I have seen posts

by some of my own Facebook friends about how they want to borrow

223

Carole A. Levitt & Mark E. Rosch, Social Media as Investigative Research, GA.

LAWS. CLE, https://georgialawyersclewebinars.ce21.com/item/social-media-investigative

-research-evidence-309411#tabDescription [https://perma.cc/78NN-FW37] (coauthors of

THE CYBERSLEUTH’S GUIDE TO THE INTERNET (14th ed. 2017)).

224

Id.

225

C. SIMON DAVIDSON, THE ETHICS OF EMAIL AND SOCIAL MEDIA: A TOP TEN LIST

302 (2017), https://www.tba.org/sites/default/files/davidson_hypotheticals_and_analysis.pdf

[https://perma.cc/QB8A-WCF2] (providing several examples of cases in which a party’s or

witness’s postings on social media sites were a useful source of evidence).

226

Id.

227

Id.

2021] #Audited: Social Media and Tax Enforcement 353

someone else’s child for their tax return, and I have seen a compilation

of tweets along the same line.

228

I presumed at least certain of these

were jokes, though sometimes it is hard to be sure. Can a data algorithm

perceive a joke?

Criminal defense attorneys advise their clients that social media is

fair game for police investigators.

229

In the civil context, divorce

lawyers routinely advise clients to be careful about what they post on

social media, cautioning that content posted can be used for a variety

of purposes.

230

Personal injury lawyers have a different set of best

practices regarding social media posts depending on whether their

client is the plaintiff or defendant.

231

A concern cited by one plaintiffs’

attorney firm is the tendency of social media posts to broadcast one’s

good news and downplay or ignore bad news, with the unintended

consequence that “defendants looking to escape liability may point to

such photos and posts as evidence that you are not really injured.”

232

For this reason, the law firm advises plaintiffs to refrain from social

media posts following an injury.

233

Harkening back to the analogy of

228

See Drumbl, supra note 69.

229

Tom Petersen, If You Are a Defendant in a Criminal Case — Be Careful What You

Post on Social Media, PETERSEN CRIM. DEF. L. (Sept. 7, 2018), https://www.criminal

defensene.com/if-you-are-a-defendant-in-a-criminal-case-be-careful-what-you-post-on

-social-media/ [https://perma.cc/XG77-S4CV].

230

See, e.g., Jaliz Maldonado, Family Law: Social Media Evidence in Divorce

Cases, THE NAT’L L. REV. (Feb. 14, 2019), https://www.natlawreview.com/article/family

-law-social-media-evidence-divorce-cases [https://perma.cc/Y2UJ-PFL8]; Larry Upshaw,

Contemplating Divorce? 10 Critical Social Media “Don’ts” You Need to Know

CONNATSER FAM. L., https://connatserfamilylaw.com/contemplating-divorce-10-critical

-social-media-donts-you-need-to-know/ [https://perma.cc/3YWS-332S] (last visited Jan.

24, 2021).

231

See, e.g., Frances Crockett Carpenter et al., Social Media Admissions in Personal

Injury Cases: Mitigating Risk for Plaintiffs, Securing Admissions from Defendants,

TRAFFORD PUBL’NS (May 10, 2017), https://www.straffordpub.com/products/social-media

-admissions-in-personal-injury-cases-mitigating-risk-for-plaintiffs-securing-admissions

-from-defendants-2017-05-10 [https://perma.cc/6WCP-YZ7K] (“This CLE webinar will

provide guidance to personal injury litigators for mitigating the risks that social media posts

pose for their clients, as well as tips for tracking down admissions by defendants on social

media.”).

232

Amalia Lucero, Social Media Being Used as Evidence in Personal Injury Cases,

CURTIS & CO. ATT’YS, https://www.curtislawfirm.org/articles/social-media-being-used-as

-evidence-in-personal-injury-cases/ [https://perma.cc/U9F3-DNAU].

233

Id. Similar advice is found on other personal injury firm websites. See, e.g.,

Adam S. Kutner, How Social Media Can Impact Your Personal Injury Case, https://

www.askadamskutner.com/personal-injury/social-media-can-impact-personal-injury-case/

[https://perma.cc/U4VK-DFGZ] (“[I]f you’re claiming that you have a broken arm, but you

post on social media that you’re going bowling, the defense is going to challenge your

injuries.”).

354 OREGON LAW REVIEW [Vol. 99, 301

other agencies that administer social benefits, an attorney representing

Social Security disability claimants states that he cautions new clients,

“There is a little bitty chance that Social Security may be snooping on

your Facebook or your Twitter account . . . . You don’t want anything

on there that shows you out playing Frisbee.”

234

What about in the tax context? Tax attorneys Carina Federico and

Travis Thompson observe, “It is imperative that modern tax

practitioners develop a full understanding of a client’s digital

footprint,” advising that “[c]lient intake questions should include

inquiries about social media usage and online sales through digital

marketplaces, at a minimum.”

235

Federico and Thompson further

suggest “practitioners should advise clients to be mindful about what

they post on social media. For instance, a taxpayer should not tell the

IRS that they do not have any money, but then post pictures of himself

on Instagram with expensive cars or on an extravagant vacation.”

236

the litigation context, tax attorney James Creech advises lawyers to

conduct due diligence of a client’s social media if the case “relies

heavily on the petitioner’s credibility on the witness stand” and warns

of the ethical consequences that may arise if a lawyer finds

contradictory evidence on social media.

237

Should low-income taxpayer clinics adopt similar routine

counseling tactics? Should our clients be counseled differently

depending on whether the case involves an examination or a collections

issue?

Beyond just the litigation context that Creech discusses, tax

attorneys should consider the wide range of their due diligence

obligations. Our clients sign financial forms such as a Collection

Information Statement or an Offer in Compromise under penalties of

perjury. The professional regulations governing practice before the IRS

impose upon tax practitioners an affirmative duty to exercise due

234

Pear, supra note 201.

235

Federico & Thompson, supra note 99, at 46.

236

Id.

237

Creech, supra note 22. Referring to Federal Rule of Evidence 803(3), Creech argues

that social media posts can be a useful indicator of mindset: “The low threshold for

publication and our cultural habit of oversharing and introspection mean that [social media

posts] are probably a fairly accurate indicator of the declarant’s mental state.” Id. While I

agree with Creech’s advice as to due diligence, it is important to remember that social media

posts are not always an accurate depiction of one’s life. See Houser & Sanders, supra note

14, at 841 (citing Minas Michikyan et al., Can You Guess Who I Am? Real, Ideal, and False

Self-Preservation on Facebook Among Emerging Adults, 3 E

MERGING ADULTHOOD 55, 60

(2015)).

2021] #Audited: Social Media and Tax Enforcement 355

diligence when preparing, assisting with, or approving tax returns and

other documents relating to tax matters, as well as due diligence in

determining the correctness of oral and written representations made

by the practitioner.

238

Before submitting such financial forms and

statements to the IRS, it would be prudent to also perform online

searches of our clients to look for inconsistencies, the way that we

might look for inconsistencies in a bank statement that supports such a

form. In the bankruptcy context, there are multiple examples of debtors

who failed to disclose assets or income and later ran into trouble

because of social media posts.

239

Elaborating on the point made by Federico and Thompson, I can

envision a number of similarly problematic tax scenarios, particularly

in the collection context. What if social media reveals a side cash

business, even one that is relatively low dollar? For our clients, that

could have an impact on the “ability to pay” calculation, and it could

materially change an Offer in Compromise. Even if the side business is

fledgling or operating at a net loss, the failure to disclose it reflects

poorly on the taxpayer’s honesty elsewhere on the form. Imagine a

client’s Facebook page posting photos of puppies for sale or advertising

to mow lawns or babysit. For low-income taxpayers in collections

cases, even one or two hundred dollars a month of unreported income

can create issues with collections alternatives, making them ineligible

for financial hardship status or forcing a higher minimum payment in

an installment agreement. This is an important conversation to have

with clients and one that can complicate calculation of monthly income

due to the unpredictability of an income stream from such sources. It

prompts a broader question as well: Should tax professionals surveil

their clients’ social media sites looking for any hint of unreported

income in connection with preparation of a routine individual income

tax return?

Another vulnerable category of taxpayers is those requesting

innocent spouse relief. In these cases, a taxpayer who has filed a joint

income tax return with his or her spouse in the past is asking for relief

from the joint liability, either because the IRS examined the return and

238

Treasury Circular No. 230, 31 C.F.R. § 10.22 (2021).

239

See, e.g., Carolyn S. Toto & Kimberly Buffington, 50 Cent Breaks the Golden Rule

of Social Media Posting, PILLSBURY INTERNET & SOC. MEDIA L. BLOG (Feb. 29, 2016),

https://www.internetandtechnologylaw.com/50-cent-breaks-the-golden-rule-of-social-media

-posting/ [https://perma.cc/55XW-HUD4]; Debtors Beware: Social Media Knows Where

Your Assets Are Buried, P

ILLSBURY INTERNET & SOC. MEDIA L. BLOG (Feb. 20, 2018),

https://www.internetandtechnologylaw.com/debtors-social-media-assets-bankruptcy/

[https://perma.cc/3ADK-4PTA].

356 OREGON LAW REVIEW [Vol. 99, 301

found an understatement of tax or because the return showed a balance

due that was not paid.

240

These cases are directed to a centralized unit

of the IRS, where an individual makes a determination using a number

of factors, including the following: whether the requesting spouse had

knowledge, or reason to know, of the understatement of tax when the

return was filed; whether the requesting spouse received a significant

benefit from the understatement or underpayment; and whether under

all the facts and circumstances it would be unfair to hold the requesting

spouse liable for the tax owed.

241

When a taxpayer requests innocent

spouse relief, the IRS sends a letter and questionnaire to that taxpayer’s

spouse or ex-spouse (the “nonrequesting spouse”) and uses that

response to evaluate the requesting spouse’s claim.

242

If the requesting

spouse appeals the determination in Tax Court, the nonrequesting

spouse has a right to intervene in the case.

243

Thus, it would be prudent

for a requesting spouse to be particularly careful about what he or she

posts to social media—both the IRS and the nonrequesting spouse may

be looking at those posts. Of course, if the requesting spouse is in the

middle of divorce proceedings, his or her divorce lawyer may already

have advised him or her to shut down social media sites. But a

significant number of innocent spouse claimants are unrepresented by

counsel,

244

particularly low-income claimants, and therefore they are

not privy to such advice.

To reiterate, I am not advising tax attorneys to help their clients

engage in tax fraud, evasion, or any kind of noncompliance. My

concerns are with broader systemic fairness. For example, will the use

of social media mining increase the likelihood that IRS employees will

make moral judgments of the poor? As Eubanks referenced in her

work, we have seen examples in other contexts in which the behavior

240

I.R.C. § 6015.

241

See I.R.C. § 6015; Rev. Proc. 2013-34, 2013-34 I.R.B. 397; I.R.S. Pub. 971 (Oct. 20,

2014).

242

Treas. Reg. § 1.6015-6(a)(1); see also IRM 25.15.3.4 (Dec. 12, 2016) (“The

[nonrequesting spouse] must receive notice of, and an opportunity to participate in, any

proceeding with respect to an innocent spouse relief request.”).

243

See King v. Comm’r, 115 T.C. 118 (2000).

244

See generally Stephanie Hunter McMahon, An Empirical Study of Innocent Spouse

Relief: Do Courts Implement Congress’s Legislative Intent?, 12 FLA. TAX REV. 629, 668

(2012) (highlighting the high number of pro se petitioners appealing the outcome of their

innocent spouse determination in court, but also remarking that “representation does not

appear to be a critical matter for determining whether a spouse wins”). Because McMahon’s

study draws upon data from cases litigated in federal courts, her statistics reflect only

requesting spouses who appealed the denial of their request in court, as opposed to the

broader universe of all requesting spouses who submit a request. Id. at 648.

2021] #Audited: Social Media and Tax Enforcement 357

of poor people is scrutinized, with officials publicly shaming the

behavior as part of a broader policy conversation.

245

I can envision a

situation in which politicians question the IRS allowance of a standard

household expense figure by citing examples of individuals who were

granted financial hardship status appearing on social media smoking a

cigarette, drinking a six-pack of beer, or sporting a tattoo. The

underlying expense of any of those things could unfortunately yet

easily become the subject of moral judgment, such as the judgment that

the taxpayer could have allocated that money to rent or food.

246

Regardless of whether the IRS chooses to move forward with social

media mining, or whether and how it publicly defines a taxpayer’s right

to privacy, it seems wise to err on the side of assuming that all

electronic transactions, actions outside one’s home, internet activity,

and social media posts may be subject to various levels of scrutiny and

observation.

247

As lawyers, whether one is a tax lawyer, a criminal

defense attorney, or a family law specialist, it is prudent to remind our

clients of this twenty-first century technological reality.

ONCLUSION

As I stated at the beginning, this Article is not meant as a defense of

those who cheat on their taxes. I certainly do not condone faking

financial hardship, fudging eligibility for social benefits, or hiding

income from the IRS. Tax compliance is a serious problem, and one

that deserves serious solutions. The underfunding of the IRS is likewise

a serious problem and compels the agency to be creative in its

enforcement solutions.

This conversation is yet another reminder that the IRS is tasked

by Congress with too many responsibilities in addition to tax

administration and revenue collection. The agency is the administrator

of refundable tax credits that operate as social benefits. It plays a

significant role in oversight of the tax aspects of retirement plans,

245

See, e.g., EUBANKS, supra note 30 (discussing how Eubanks uses the example of the

public shaming of TANF recipients who use ATMs in certain locations).

246

See generally Zelenak, supra note 56 (discussing how the public views tax cheats

versus welfare cheats).

247

See, e.g., Heather Kelly & Rachel Lerman, America Is Awash in Cameras, a Double-

Edged Sword for Protesters and Police, WASH. POST (June 3, 2020, 4:00 AM), https://

www.washingtonpost.com/technology/2020/06/03/cameras-surveillance-police-protesters/

[https://perma.cc/53ZG-9YKA] (describing law enforcement’s use of surveillance cameras,

body cameras, and face- and object-recognition software, as well as private citizens’ use of

smart phones, home security cameras, and vehicle cameras, and describing the metadata

contained in these videos and photos).

358 OREGON LAW REVIEW [Vol. 99, 301

nonprofit organizations, and healthcare-related provisions. It gets

called upon to administer financial aid in times of crisis, as in the

COVID-19 pandemic when the IRS was unexpectedly tasked with

delivering economic impact payments to millions of individuals,

including those with no taxable income or filing requirement. It stands

to reason that the agency—overtasked and underfunded—is searching

for shortcuts in the name of efficiency.

Despite these pressure points, to which I am sympathetic, I sincerely

hope that the agency will choose not go down the road of cheap and

easy in relying on big data and social media analytics for collection and

examination purposes. The IRS aspires in its mission statement to

“enforc[e] the tax law with integrity and fairness to all.”

248

In my view,

incorporating social media mining as a routine part of examination and

collection would undermine the dignity of the taxpayers, as well as the

integrity of the agency.

248

I.R.S. Policy Statement 1-236 (Oct. 24, 2016).